core.ac.uk · 2015-09-22 · uwoliov s%(> STELLINGEN 1. De 'omvattende reconstructie', het streven...
Transcript of core.ac.uk · 2015-09-22 · uwoliov s%(> STELLINGEN 1. De 'omvattende reconstructie', het streven...
THE REGULATION OF THE STRUCTURE AND FUNCTION OF FLAVIN (VITAMIN B2) ON BINDING TO APOFLAVOPROTEINS AND ITS BIOLOGICAL IMPLICATIONS
CENTRALE LANDBOUWCATALOGUS
0000 0086 5689
Promotor: dr. F. MUI 1er
hoogleraar in de biochemie
Chrit Moonen
THE REGULATION OF THE STRUCTURE AND FUNCTION OF FLAVIN (VITAMIN B2) ON BINDING TO APOFLAVOPROTEINS AND ITS BIOLOGICAL IMPLICATIONS
Proefschrift
ter verkrijging van de graad van
doctor in de landbouwwetenschappen,
op gezag van de rector magnificus,
dr. C.C. Oosterlee,
in het openbaar te verdedigen
op vrijdag 16 december 1983
des namiddags te vier uur in de aula
van de Landbouwhogeschool te Wageningen.
IS*/* \%Uoo-°l
uwoliov s%(>
STELLINGEN
1. De 'omvattende reconstructie', het streven naar 'numeriek rendement'
en 'modulegewijze opbouw' voor het geplande wetenschappelijk onderwijs
(concept beleidsnota Beiaard) ontkent het wetenschappelijk onderwijs als
een culturele waarde, die onlosmakelijk verbonden is met de identiteit
van onze cultuur.
2. De individuele vrijheid, die als een normatieve kracht in het denken te
voorschijn treedt, is een fundamentele verworvenheid in onze cultuur. In
dit licht moeten we het afstemmen van het wetenschappelijk onderwijs op
de arbeidsmarkt als een inbreuk op de individuele vrijheid zien.
3. Nu de invloed van de religie in Nederland als leverancier van normen en
waarden steeds minder wordt, is het van des te groter belang om aandacht
te schenken aan de zin en de ontwikkeling van normen en waarden.
4. In discussies over de legitimatie van wetenschappelijk onderzoek spelen
de kwalificaties 'toegepast' of 'fundamenteel' onderzoek een belangrijke
rol (zie bijv. de nota 'Biochemie Over Leven'). Deze kwalificaties dienen
achterwege gelaten te worden, zolang deze terminologie gebrekkig gedefi
nieerd en daardoor subjectief is.
'Biochemie Over Leven', Staatsuitgeverij, 's-Gravenhage, 1982.
5. In het zoeken naar een verklaring voor het 'afstemmechanisme' van de func
tie van flavine co-enzymen door apoflavoproteinen is de rol van de modulatie
van de aktiveringsenergie in de verschillende redoxovergangen onderschat.
6. Aangezien röntgenkristallografische gegevens van gereduceerde flavoëiwitten
wat betreft het isoalloxazine passend worden gemaakt met slechts één vari
abele, i.e. de 'vlinderhoek', zijn de conclusies omtrent de werkelijke con-
formatie van de isoalloxazinering twijfelachtig.
7. De algemene aanvaarding van het C(4a) peroxyflavine als een essentieel in
termediair in de activatie van zuurstof door flavines is voor een belangrijk
deel te danken aan het feit, dat deze verbinding door Ghisla et_ al. d.m.v. 13
C NMR is waargenomen in het enzym luciferase. De door de auteurs aan het
C(4a) peroxyflavine toegekende piek representeert dit intermediair echter niet.
Ghisla, C., Hastings, J.W., Favaudon, V. and Lhoste, J.-M.
Proc.Natl.Acad.Sci. USA 75, 5860-5863, 1978.
8. Hoewel reeds lang bekend is, dat C(4a) peroxyflavine een geschikt hydroxy-
leringsreagens is voor electronenrijke verbindingen, concludeert Visser
uit enkel de structuur van het geoxydeerde enzym-substraat complex van
p-hydroxybenzoaathydroxylase, dat het enzymatische reactiemechanisme geen
analogon in de organische chemie kan hebben. De omgekeerde conclusie i.e.
de reactie heeft een analogon en daarom is het model onjuist, moet waar
schijnlijker worden geacht.
Visser, C.M., Eur.J.Biochem. 135, 543-548, 1983
9. De resultaten van Tauscher et al., overigens grotendeels onjuist, hebben
voor het onderzoek naar de modulatie van redoxpotentialen van eiwitgebon
den flavines een werking gehad, die in biochemische termen omschreven kan
worden als een 'dead-end' of 'suicide' inhibitie.
Tauscher, L., Ghisla, S. and Hemmerich, P., Helv.Chim.Acta 56,
630-649, 1973.
10. Schakende promovendi moeten zich realiseren dat hun stelling voor een
opponent wel eens mat zou kunnen zijn.
Chrit Moonen
The regulation of the structure and
function of flavin (vitamin B_) on
binding to apoflavoproteins and its
biological implications
16 december 1983, Wageningen
LANDBOUWHOGESCHOOI WAGKNfNGKlV
"With the role of flavins in bacterial bioluminescence as precedent, we can
anticipate that flavins may continue to provide their own spotlight on their
roles in redox biochemistry"
Christopher Walsh
The work described in this thesis has in part been carried out under the
auspices of the Netherlands Foundation for Chemical Research (SON) with
financial aid from the Netherlands Organization for the Advancement of
Pure Research (ZWO).
L A N D B O U W H O G E S C H O O L W A G E N I N G E N
Voorwoord
Ondanks het (te) individualistische karakter van het promotieonderzoek,
kan ik terugzien op perioden van uitstekende samenwerking, waarvoor ik met
genoegen mijn dank uitspreek. Bij voorbaat excuseer ik mij, indien ik mensen
mocht vergeten.
Mijn promotor, Franz Muller, voor het eindeloze geduld, waarmee hij
mijn ideeën aanhoorde en corrigeerde. Hij betekende meer voor mij, dan de
term „promotor" aanduidt.
Willem van Berkel en Willy van den Berg isoleerden en analyzeerden voor
mij de eiwitten, zodat ze aan de belangrijkste criteria voldeden: 1) het
moet geel zijn; 2) het moet veel zijn.
Jacques Vervoort en Robert Wijnands, die mij met allerhande werkzaam
heden hebben geholpen én mede de ontspannen sfeer op lab VI bepaalden.
Marleen Boerjan en Sjra Maessen, die bij mij een onderzoek hebben ver
richt in het kader van hun studie Moleculaire Wetenschappen.
Chris Rasmussen voor de gedegen hulp bij de evaluatie van de resultaten
en het corrigeren van de uiteindelijke tekst.
Hans Bosma, Aart de Kok en Cees Veeger voor de diskussies over lipoamide
dehydrogenase.
Johan van Leeuwen voor zijn gedegen kritiek, zodra er sprake was van
ladingen.
Jenny Toppenberg-Fang en Lyda Verstege voor het keurige typewerk.
Martin Boumans en Bery Sachteleben voor het tekenwerk.
Dit onderzoek zou vrijwel onmogelijk geweest zijn zonder NMR machines
van uitstekende kwaliteit. Van onschatbare waarde hierbij waren Adrie de
Jager voor de Bruker CXP 300 en Varian XL100, Klaas Dijkstra voor de Bruker
HX360, Pieter van Dael en Cees Haasnoot voor de Bruker WM500 en Joost Lohman
voor de Bruker WM250.
Niet alleen de Groote Markt zorgde ervoor, dat ik graag naar Groningen
ging. De vakgroep Fysische Chemie zorgde steeds voor een goede sfeer en
grote behulpzaamheid. Voor mij fungeerden deze bezoeken als een soort bij-
scholing in NMR. Daarvoor mijn hartelijke dank aan Ruud Scheek, Rolf Boelens,
Sytze Stob, Peter Hore, Klaas Dijkstra en Rob Kaptein.
Chapters 2,3,4 have been published separately.
Chapters 5,6,7,8,10,11 will be published separately.
Chapter 2: C.T.W. Moonen and F. Müller, Biochemistry, 1982, 21, 408-414. Chapter 3: C.T.W. Moonen, P.J. Höre, F. Müller, R. Kaotein and" S.G. Mayhew,
FEBS Letters, 1983, U9_, 141-146. Chapter 4: C.T.W. Moonen and F. Müller, Eur.J.Biochem., 1983, JL33, 463-470. Chapter 5: C.T.W. Moonen, J. Vervoort and F. Müller, submitted to Biochemistry. Chapter 6: C.T.W. Moonen, J. Vervoort and F. Müller, submitted to Biochemistry. Chapter 7: C.T.W. Moonen and F. Müller, submitted to Eur. J. Biochem. Chapter 8: C.T.W. Moonen, W.A.M. van den Berg, M. Boerjan and F. Müller,
submitted to Biochemistry. Chapter 10: C.T.W. Moonen and F. Müller, submitted to Eur. J. Biochem. Chapter 11: C.T.W. Moonen, R.M. Scheek, R. Boelens and F. Müller, submitted to
Biochemistry..
Contents Page
Chapter 1
Chapter 2
Chapter 3
Chapter 4
Chapter 5
Chapter 6
Chapter 7
Chapter 8
Chapter 9
Chapter 10
Chapter 11
General introduction:
At the crossroad of one- and two-electron redox bio
chemistry: the flavin coenzymes.
Structural and dynamic information on the complex of
Megasphaera elsdenii apoflavodoxin and riboflavin 5'-
phosphate. A Phosphorus-31 Nuclear Magnetic Resonance
study.
A photo-CIDNP study of the active sites of Megasphaera
elsdenii and Clostridium MP flavodoxins.
On the mobility of riboflavin 5'-phosphate in Megasphaera 13 elsdenii flavodoxin by C-Nuclear-Magnetic-Resonance
Relaxation.
A reinvestigation of the structure of oxidized and re-13 15
duced flavin. A C and N Nuclear Magnetic Resonance
study.
A Carbon-13 Nuclear Magnetic Resonance study on the dyna
mics of the conformation of reduced flavin.
On the intermolecular electron transfer between different
redox states of flavodoxin from Megasphaera elsdenii: A
500 MHz *H NMR study. 13 15
A C and N Nuclear Magnetic Resonance study on the interaction between riboflavin and Riboflavin Binding Protein. The different classes of flavoproteins: Miscellaneous NMR results and mechanistic implications. A proton Nuclear Magnetic Resonance study at 500 MHz on Megasphaera elsdenii flavodoxin.
The use of two-dimensional NMR spectroscopy and two-dimen
sional difference spectra in the elucidation of the active
center of M.elsdenii flavodoxin. Summary
Samenvatting
List of abbreviations
Curriculum Vitae
15
21
29
57
70
92
112
142
166
184
188
191
192
1 -
Chapter 1: General Introduction
At the crossroad of one- and two-electron redox biochemistry: the flavin
coenzymes.
The term „flavin" refers to the yellow chromophore of a class of respi
ratory enzymes, occurring widely in animals and plants, i.e. the flavoproteins.
The yellow chromophore is the tricyclic isoalloxazine molecule. Although va
rious flavin coenzymes exist (for example, riboflavin (vitamin B - ) , FMN, FAD)
the isoalloxazine ring system is the functional part of the prosthetic group.
In Scheme 1 the most common flavin coenzymes are shown including the interna
tionally accepted numbering system.
photolysis ©
x CO o
X OJ
o
3
Z"1 5 z
V 0
lumichrome
lumiflavin
pH<7 pH>9 0 0
II H 2
CH 2 — (CH0H)3 — CH20 — P — O - j - P — 0 — C
© © H or OH
riboflavin (vitamin B?)
I OH
riboflavin - 5 - monophosphate (FMN)
flavin - adenine - dinucleotide (FAD)
OH ^
r, IN
N V N H -
•a' OH OH
Scheme 1
The most prominent features of the biochemical properties of flavin are
the redox properties. As indicated in the title of this chapter, flavins can
undergo both facile two- and one-electron redox transitions, which makes fla-
vins unique among all known redox coenzymes. The basis of this lies in the
chemical properties of flavin as outlined in Scheme 2. The possible kationic
and anionic forms of each state are added in Scheme 2, but it should be noted
that only the neutral and anionic species are of biological relevance.
flavoquinone
H3C>
H3C-©
FI0XH2® 0
flavohydroquinone R
H 3 C ^ Y N y N Y ° PKa<0 a H
FlredHi©
3' FlrecjH3 Flr edH2©
Scheme 2
Another prominent feature of flavin coenzymes is the high reactivity for
reductive oxygen metabolism. This property is used in many different reactions
catalyzed by flavoproteins. Especially mentioned is the light generating reac
tion catalyzed by luciferase, to which the remarkable statement of Cristopher
Walsh refers on the first page of this dissertation.
Another feature distinguishes flavin coenzymes from, for example, nico
tinamide coenzymes. This is the fact, that flavin coenzymes are always complexed
with apoproteins. Thus flavin forms very strong, but generally non-covalent,
complexes with apoflavoproteins. Since the flavin possesses a different elec
tronic structure and possibly also a different conformation in the three redox
states, the affinity of the apoprotein for flavin generally differs for the
three redox states. This implies that the redox potential of the protein-bound
flavin is modified by the binding mode of a particular flavoprotein.
It is not surprising that flavin is often used in nature in a wide variety
of biochemical reactions, since the required chemical properties are combined
in one molecule. In Table 1 a classification of the various flavoproteins is
given on the basis of the nature of the electron donor and acceptor. Other
classifications have been proposed which are not presented because of their
complexity and of doubtful use in biochemistry.
Table 1. A possible classification of flavoproteins.
Category Comments Example
1) Dehydrogenases
2) Oxidases
3) Oxidases-decarboxylases
4) Monooxygenases (hydroxylases)
5) Dioxygenases
6) Metalloflavo-proteins
7) Flavodoxins
Acceptors are often 1 e" acceptors such as cytochromes or nonheme-iron-sulfur clusters
Acceptor is 09, which is reduced by 2e" to H ^
Acceptor is 0 ? , which is reduced by 4e" to H20
Acceptor is 0~: one oxygen ends up in H~0 and the other in the the hydroxylated product
Acceptor is 0«; both oxygen atoms end up in the product
Acceptor may be Op.jÇoundjtpansi-tioo-metal ions (Fe , Fe or Mo ) are required. May be dehydrogenases or oxidases
le" transfer proteins. Acceptor may be nitrogenase
lipoamide dehydrogenase
D-amino acid oxidase
L-lactate oxidase
p-hydroxybenzoate hydroxylase
2 nitropropane dioxygenase
xanthine oxidase
flavodoxin
For more detailed information the reader is referred to some recent review
articles (Walsh, 1980; Müller, 1982; Brui ce, 1980; Massey and Hemmerich, 1980),
to the introduction of the following chapters and to the references given there
in.
- 4 -
The striking versatility of flavin bound to different apoflavoproteins has
led to the hypothesis that upon binding to apoflavoproteins the chemical and
physical properties are modified in such a way that the coenzyme is „tuned" to
its actual function in a particular flavoprotein (Müller, 1972). Although this
hypothesis appears to be generally accepted at the moment, relatively little
is known about the actual mechanism. The present knowledge stems mainly from
kinetic data on flavoproteins (see Massey and Hemmerich, 1980 and references
therein), model studies (Bruice, 1980) and crystallographic data on a few
flavoproteins (Wierenga et a]_., 1979 and references therein). How impressive
these studies may be, they do not offer yet adequate insight to explain the
observed properties of flavoproteins, as can be deduced from the following,
i) From the chemistry of model compounds it is learned that the reactivity of
flavins changes with a change in electron density at a particular atom in the
flavin. However, nothing is known about how these changes in electron density
are induced in flavoproteins.
ii) Crystallographic studies revealed the detailed structure of, for example,
flavodoxin from Clostridium MP (Ludwig et al_., 1976) and p-hydroxybenzoate
hydroxylase from Pseudomonas fluorescens (Wierenga et al_., 1979). However,
ambiguities still exist in the action mechanisms of both proteins (Chapters
7 and 9).
iii) Although it is known that redox potential modulation is a very important
feature of apoflavoproteins (Walsh, 1980), the mechanism is still badly under
stood. Simondsen and Toll in (1980) argued that conformational effects exerted
by the apoflavoprotein on the flavin are the main factors governing the redox
potential modulation. This model is, however, too simplified as shown in Chap
ter 9.
Therefore, to arrive at a more detailed explanation of the tuning mechanism
of protein-bound flavin, a method is required allowing a detailed insight into
the molecular and submolecular structure of flavin, i.e. to obtain information
on changes in hydrogen bond formation and electron densities. Two suitable
techniques are available at the moment, i.e., Raman spectroscopy and Nuclear
- 5 -
Magnetic Resonance (NMR) spectroscopy. Although promising, Raman data appear
at the moment to be difficult to interpret (Müller et aj_., 1983). The NMR tech
nique is the most versatile technique and allows the observation of ir electron
density at specific carbon and nitrogen atoms of the flavin ring.
In addition, phosphorus and proton NMR can be used to monitor the structure
of the flavin side chain and of the protein, respectively. Moreover, dynamic
information can be obtained from NMR measurements. NMR studies on biomolecules
are, however, hampered by the fact, that the method is in principle an in
sensitive method. Tremendous advances in NMR methodology and instrumentation
in the last decade have brought the method closer to a meaningful application
to biochemical problems. Unfortunately, natural abundance carbon and nitrogen
13 15
contain only 1.1% and 0.4% of the NMR sensitive C and N isotope, respec
tively. This problem can be circumvented by isotopic enrichment, which is often
used in such studies. For an introduction into the NMR technique the reader is
referred to textbooks of Wüthrich (1976), Jardetzky and Roberts (1981) and
Dwek (1973).
The aim of this research project was to investigate the particular mecha
nism of tuning the properties of the flavin coenzyme by apoflavoproteins. Two
different approaches were used in this thesis.
1) A study of the electronic and conformational structure of the protein-bound
flavin.
2) A study of the protein to investigate the interaction between the flavin
coenzyme and a particular apoflavoprotein.
ad 1. The basis of this approach lies in the chemical synthesis of coenzymes in
which certain carbon or nitrogen atoms are replaced („enriched") by the 13 15 C or N isotopes, respectively. After removing the natural coenzyme,
the protein can be reconstituted with the „enriched" coenzyme. In this
way one obtains a flavoprotein, which does not differ in any chemical
properties from the original flavoprotein, but which contains a built-in
13 15 monitor in the form of the particular C or N atom. This approach allows to obtain a detailed view of one single atom of the flavin coenzyme
- 6 -
in the various redox states. In some favorable cases even the changes
of the atom during the catalytic cycle can be followed.
The specific information obtained in this way must be „translated" into
hydrogen bonds, structure, electron density and mobility. The data ob
tained on free flavin resulted in a semi-empirical theory which seems
to be a valuable tool to interpret data obtained on flavoproteins.
31 ad 2. P NMR is a relatively simple method, which gives direct information
on the interaction of apoflavoprotein and the phosphate group(s) of the
31 1 flavin coenzyme. Apart from P NMR, H NMR is a extremely powerful
method for the elucidation of the structure of flavoproteins and for a
description of the interaction between flavin and apoflavoprotein. How
ever, H NMR suffers from the lack of resolution as the flavoproteins
contain an overwhelming number of hydrogen atoms. However, the recent,
fast developing techniques in the field of H NMR of biomolecules are
very promising for an application to the smallest flavoproteins, the
flavodoxins. In this thesis especially time-resolved photochemically
induced dynamic nuclear polarization (CIDNP) and two-dimensional H NMR
spectroscopy were applied. The problem of the resolution was partially
circumvented by making use of the paramagnetic semi qui none state. The
information, collected by these methods is very specific and approaches
information obtained by x-ray crystallography.
This thesis respresents by no means an exhaustion of the potentialities
of the method in the field of flavoprotein biochemistry. On the contrary, from
the still advancing methods and instruments it can be expected that the Nuclear
Magnetic Resonance technique will continue to offer new insights into this
field.
- 7 -
Bruice, T.C. (1980) Acc.Chem.Res. 13, 256-262.
Dwek, R.A. (1973) Nuclear Magnetic Resonance (NMR) in Biochemistry",
Clarendon Press, Oxford.
Jardetzky, 0. and Roberts, G.C.K. (1981) „NMR in Molecular Biology",
Academic Press, New York.
Ludwig, M.L., Burnett, R.M., Darling, G.D., Jordan, S.R., Kendall, D.S. and
Smith, W.W. (1976) in: „Flavins and Flavoproteins" (Singer, T.P., Ed.),
Elsevier, Amsterdam, 393-404.
Massey, V. and Hemmerich, P. (1980) Biochem.Rev., 8, 246-257.
Müller, F. (1972) Z.Naturfors. 27b, 1023-1026.
Müller, F. (1983) Topics in Current Chemistry 108, 71-107
Müller, F., Vervoort, J., Lee, J., Horowitz, M. and Carreira, L.A. (1983)
J.Raman Spectr. 14, 106-117.
Simondsen, R.P. and Tollin, G. (1980) Mol.Cell.Biochem. 33, 13-24.
Walsh, C. (1980) Acc.Chem.Res. 13, 148-155.
Wierenga, R.K., De Jong, R.J., Kalk, K.H., Hol, W.G.J. and Drenth, J.J. (1979)
J.Mol.Biol. 131, 55-73.
Wütrich, K. (1976) „NMR in Biological Research: Peptides and Proteins"
North Holland Publishing Company, Amsterdam.
- 8 -
408 Biochemistry 1982, 21, 408-414
Chapter 2
Structural and Dynamic Information on the Complex of Megasphaera elsdenii Apoflavodoxin and Riboflavin 5'-Phosphate. A Phosphorus-31 Nuclear Magnetic Resonance Study^ Chrit T. W. Moonen and Franz Müller'
ABSTRACT: It is shown that commercial FMN contains a considerable amount of the 4', 3', and 2' isomers and other phosphorus-containing compounds. These impurities can easily be analyzed and quantified by the 31P NMR technique. The phosphate group of FMN bound to Megasphaera elsdenii apoflavodoxin is probably in the dianionic form. Its chemical shift is almost independent of pH in the range 5.5-9.2 and of the redox state of the protein. The phosphate group of bound FMN is buried in the protein. Protons of the apoprotein and of bound water located in the vicinity of the phosphate group in native flavodoxin are not exchangeable with deuteron of the bulk solvent. These protons are, however, easily exchangeable in apoflavodoxin, as shown by reconstitution experiments. The distance between the phosphorus atom of bound FMN and
the N(10) atom of the isoalloxazine moiety of FMN is calculated to be about 7.8 Â. This result is in good agreement with X-ray data published for the related flavodoxin from Clostridium MP. The electron exchange between the oxidized arid semiquinone state of M. elsdenii flavodoxin is rather slow (T »0.5 s) whereas that between the semiquinone and
nyüroquinone form is much more favored (T <<0.01 s) This indicates that the activation energy lor tne transition between the semiquinone and hydroquinone states must be smaller than that for the transition between the oxidized and semiquinone states. These results offer a reasonable explanation for the one-electron transfer reaction of flavodoxins in biological reactions.
J 7 lavodoxins are proteins of small relative molecular mass functioning as electron carriers in low potential oxidation-reduction reactions (Mayhew & Ludwig, 1975). In these redox reactions the flavodoxins are often interchangeable with the iron-sulfur proteins ferredoxins. The flavodoxins contain riboflavin 5'-phosphate (FMN)1 as prosthetic group. Flavodoxins can be reduced in two distinct one-electron steps with the formation of the relatively oxygen-stable flavosemiquinone as an intermediate (Mayhew & Massey, 1973). During electron transfer reactions, flavodoxins shuttle between the semiquinone and hydroquinone state (Mayhew & Ludwig, 1975, and references therein). The chemical and physical properties of the flavodoxins have been investigated in great detail (Mayhew & Ludwig, 1975).
Flavodoxin from Megasphaera elsdenii is easily available in large quantities. Its physical chemical properties are very similar to those of Clostridium MP flavodoxin for which three-dimensional structural data at high resolution are published (Smith et al., 1977). Because the latter flavodoxin is much more difficult to isolate in the large quantities required for certain physical techniques (e.g., nuclear magnetic resonance), M. elsdenii flavodoxin is often used as a substitute in such studies. In the past the binding of FMN to apoflavodoxin of M. elsdenii has been investigated by fluorescence and visible light absorption techniques (Mayhew & Massey, 1969) and by conventional kinetic (Gast et al., 1976) and temperature-jump techniques (Gast & Müller, 1978). In order to complement these studies and to gain, if possible, insight into the interactions between the apoenzyme and its prosthetic group on an atomic level, we set up a NMR program on M. elsdenii flavodoxin and related biomolecules. The aim of such studies is to elucidate the subtle, specific interactions between the constituents of a flavoprotein which are probably, as proposed
* From the Department of Biochemistry, Agricultural University, De Dreijen 11, 6703 BC Wageningen, The Netherlands. Received July 1, 1981. This work was supported by the Netherlands Foundation for Chemical Research (SON) with financial aid from the Netherlands Organization for the Advancement of Pure Research (ZWO).
earlier (Müller, 1972), responsible for the specific biological action of a certain flavoprotein.
Here we report on a 31P NMR study on the binding of FMN by M. elsdenii apoflavodoxin. The results yielded information on the interaction of the phosphate group of the prosthetic group with the apoprotein, the distance between the phosphorus atom and the isoalloxazine moiety, and the rate of electron exchange between the molecules in the different redox states. In addition, it is shown that commercial FMN contains several isomeric compounds which can be identified by their chemical shifts and quantitated by the integrals of the corresponding resonance lines.
Materials and Methods Flavodoxin from Megasphaera elsdenii was isolated and
purified according to published procedures (Mayhew & Massey, 1969). FMN was obtained from Sigma (lot 44B-0820). 7,8-Dimethyl-Af10-(2-hydroxyethyl)isoalloxazine 2'-phosphate was synthesized as described earlier (Müller et al., 1973). Sodium dithionite was purchased from Merck, Darmstadt, West Germany.
For removal of the exchangeable protons in the "interior" of the flavodoxin, the apoflavodoxin was prepared according to the method of Wassink & Mayhew (1975). The apoflavodoxin was then dissolved in a minimal volume of a solution of 50 mM Tris-HCl in deuterium oxide (pH 8.0). This solution was dialyzed against the same buffer for 1 day at room temperature and then for 2 days at 4 °C. The buffer solution was exchanged 2 times against freshly prepared deuterated buffer solutions. After the last change, a 3-fold excess of FMN was added to the fresh buffer and equilibrated for 1 day at 4 °C. Excess of FMN was removed from reconstituted flavodoxin by dialysis against a solution of 50 mM Tris-HCl in H20 (pH 8.0). The flavodoxin thus obtained was then concentrated by lyophilization.
1 Abbreviations: FMN, riboflavin 5'-phosphate; 4'-FMN, riboflavin 4'-phosphate; 3'-FMN, riboflavin 3'-phosphate; 2'-FMN, riboflavin 2'-phosphate; NMR, nuclear magnetic resonance; EDTA, ethylenedi-aminetetraacetic acid; Tris, tris(hydroxymethyl)aminomethane.
Reprinted from Biochemistry, 1982, 21, 408. Copyright © 1982 by the American Chemical Society and reprinted by permission of the copyright owner.
- 9 -
"P NMR OF M. E L S D E N I I F L A V O D O X I N
Reduction and reoxidation experiments were conducted by the addition of the desired amount of a dithionite solution to the anaerobic solution of flavodoxins or free FMN. Anae-robiosis was achieved by carefully flushing the solutions in the NMR tube with argon for about 20 min. The desired degree of reoxidation was obtained by injecting small volumes of air into the NMR tube containing the anaerobic flavohydro-quinone solution followed by gentle mixing.
For relaxation measurements on solutions of free FMN, traces of paramagnetic metal ions were removed from the solutions by passing them through a small Chelex-100 (product of Bio-Rad) column. In all other cases a small amount of EDTA was added to the solution to be investigated.
31P NMR measurements were recorded on a Varian XL 100-15 spectrometer operating at 40.5 MHz and equipped with a 16K Varian 620-L computer. A spectral width of 1000 Hz (4K data points) and an acquisition time of 2 s was used. All spectra were acquired under proton noise decoupling conditions unless otherwise stated. All chemical shift values were determined relative to the external standard 85% H3P04. The spectrometer was locked on the deuterium resonance line of deuterium oxide contained in the buffer solution. All spectra were recorded at 26 ± 2 °C.
Wilmad 12-mm precision NMR tubes were used. The samples contained 1.5-3 mM flavodoxin in 150 mM Tris-HCl solutions of pH 8.0. Samples of free FMN contained 56 mM FMN in the same buffer (pH 8.2) unless otherwise stated. Depending on the kind of experiment, the buffer solutions were made up of 10-100% deuterium oxide. In pH titration experiments the desired pH value was adjusted in the NMR tube by addition of 0.1 M HCl or solid Tris. The pH meter readings were used without correction for isotope effects.
Spin-lattice (T:) relaxation measurements were performed by using the inversion-recovery method (Void et al., 1968) as described by Levy & Peat (1975). At least 10 spectra were recorded for the determination of one 7", value. The experimental values were fitted to one exponential function by computer analysis (Sass & Ziessow, 1977).
Spin-spin (T2) relaxation values were estimated from the line width of the experimental spectra.
Equilibrium ultracentrifugation was performed in a MSE analytical ultracentrifuge. Samples contained 0.5-3 mM flavodoxin in 150 mM Tris buffer of pH 8.0. Measurements were done at 26 °C. These conditions are the same as used in the NMR experiments.
Results and Discussion "P NMR of Free FMN. It has been shown that apofla-
vodoxin from M. elsdenii binds only those flavin derivatives which carry at the N(10) position a side chain of five carbon atoms and a terminal phosphate group (Mayhew, 1971; Gast & Müller, 1978) (Figure 1). Mayhew (1971) and Wassink & Mayhew (1975) demonstrated from binding studies of apoflavodoxin from M. elsdenii that commercial FMN contains impurities (20-30%) not binding to the apoflavodoxin. Pure FMN, however, can be obtained by affinity column chromatography using apoflavodoxin from M. elsdenii as an affinity label (Mayhew & Strating, 1975). More recently Scola-Nagelschneider & Hemmerich (1976) demonstrated from 'H NMR data that riboflavin 4'-phosphate (4'-FMN) is the major byproduct of FMN purified by ion-exchange chromatography. It is known that the chemical shift of phosphate esters directly reflects the bond angles in the O-P-O grouping (Gorenstein, 1975). These bond angles are also dependent on the ionization state of the phosphate group in phosphate esters such as FMN. In addition hydrogen bonds
V O L . 2 1 , N O . 2 , 1982 409
5' CH 2 0P0 3 H 2
FIGURE 1: Structure of riboflavin 5'-phosphate (FMN).
jJlLli-L_z
- 6 - 5 'U - 3 - 2 PPM(Ó)
0
FIGURE 2: 3,P NMR spectrum of commercial FMN (56 mM) in 150 mM Tris-HCl, pH 7.0.
and stereochemical effects also affect the 31P chemical shifts, but to a lesser degree than the bond angle (Evans & Kaplan, 1979). The polarity of the environment of the phosphate ester does not, or at most only to a minor degree, influence the 31P chemical shifts (Gorenstein et al., 1976). These facts make it possible to interpret 31P NMR spectra accurately. Therefore, commercial FMN was also investigated with the aim to identify, if possible, the byproducts.
In Figure 2 the 31P NMR spectrum of commercial FMN is shown. The pH value used here (pH 7.0) gives the best resolution of the resonance lines. From this spectrum it is obvious that besides FMN and 4'-FMN, which are the major components of the sample, other phosphorus-containing compounds are present (Table I). For characterization and facilitation of the assignment of the resonance lines of the major components in Figure 2, a pH titration study was performed. The minor components in Figure 2 (peaks 5-7) were not analyzed further. It was found that the phosphate groups exhibit a pK value of about 6 and that all four resonances in Figure 2 (peaks 1-4) show a downfield shift of 3.5-5 ppm on going from pH 4.0 to 9.0. This strongly indicates that all major resonances are due to phosphate esters. The small differences in their pKt values reflect probably some stereochemical differences among these esters as is also obvious from the corresponding chemical shift values (Table I). For instance, the pATa value of the phosphate ester corresponding to peak 4 of Figure 2 is higher than that of FMN (peak 2). This may be due to a weak interaction of the OH group of the phosphate ester with the isoalloxazine ring, i.e., N(l). To check this possibility the model compound 7,8-dimethyl-7V'°-(2-hydroxyethyl)isoalloxazine 2'-phosphate was studied. This compound indeed shows a pATa value identical with the phosphate ester at -3.17 ppm (Table I). Under identical conditions the chemical shift of the model compound is about
10 -
410 BIOCHEMISTRY MOONEN AND MULLER
Table I: Analysis of Commercial FMN by 31
peak in Figure 2
1 2 3 4 HIP"
Ptfa"
5.7 6.1 6.1
-6.3 d
6.3
abundance (%)6
~5 78 ~9 ~4
PNMR
chemical shifts c at pH
4.0
-0.15 -0.90 -0.07
0.20 0.10
7.0
-4.48 -4.01 -3.65 -3.17 -2.85
9.0
-4.92 -4.70 -4.33 -3.90 -3.83
assignment
3'-FMN FMN 4'-FMN 2-FM.N
a The accuracy of the values is ±0.05 pH unit. The pKa values were calculated by fitting pH vs. chemical shift curves. b Quantitatively determined by integration of 31P NMR spectra where the delay between two accumulations was greater than 5Tr No exponential multiplication was applied in such experiments. c Chemical shifts (ppm) are reported relative to external H3P04 (85%). d This value could not be determined with the same accuracy as the other reported values owing to the overlap with other peaks. e 7,8-Dimethyl-A'"1-(2-hydroxyethyi)isoalloxazine 2'-phosphate.
0.3 ppm toward higher field. The downfield shift of the corresponding phosphate group in 2'-FMN is explained by the difference in substitution of the 2' position in the two molecules. On the basis of these results, peak 4 (Figure 2) can be assigned with great confidence to a 2'-phosphate ester of riboflavin.
The assignments of the resonance lines of the 3IP NMR spectrum of commercial FMN (Figure 2) as presented in Table I were confirmed by investigating a solution of half-reduced FMN. It is known that such solutions contain about 4% flavosemiquinone at pH 7 (Müller et al., 1971). This intramolecular paramagnetic label should yield information with respect to the distance between the various phosphate esters and the isoalloxazine ring system by the mechanism of dipolar broadening of the resonance lines. We observed that in spectra of such solutions peak 1 and peak 4 were no longer detectable because of broadening, and peak 2 was hardly broadened, whereas peak 3 showed a greater degree of broadening than peak 2. This indicates that the distance between the phosphate ester group and the isoalloxazine ring system increases in the series: peaks 4 and 1, peak 3, and peak 2. These results support the assignments given in Table I.
The 31P chemical shift of FMN reported in this paper is in agreement with values published by Favaudon et al. (1980) and Edmondson & James (1979). In addition it was found that NaCl and urea in concentrations up to 1 M did not influence the 31P chemical shift of FMN (pH 8.2). This is in accordance with the conclusion of Gorenstein et al. (1976) from results on other phosphate esters that 31P chemical shifts are only little influenced by the environment.
It is known that FMN in aqueous solution aggregates; i.e., stacking of the isoalloxazine ring system occurs. Sarma et al. (1968) investigated this phenomenon by 'H NMR. We extended this study with 'H NMR to lower concentrations than previously used. Our results are fully consistent with those of Sarma et al. (1968). It was, however, found that stacking of FMN already occurs at concentrations as low as 0.3 mM (100 mM KPj, pH 6.5). In fact it was demonstrated by light absorption difference spectroscopy that monomeric FMN is only found at concentrations lower than 50 nM (Müller et al., 1973). Moreover 31P T, measurements indicate that the mobility of the ribityl phosphate side chain of FMN decreases with increasing concentrations of FMN, i.e., increasing stacking. For concentrations of 3.5, 14.2, and 59.3 mM FMN in 50 mM Tris, pH 8.2, we found the following Tt values: 4.7, 4.3 and 3.9 s, respectively. These T{ values are in the same range as those of adenosine 5'-phosphate measured at 40.5 MHz (Nanda et al., 1980).
FIGURE 3: "P NMR spectra of M. elsdenii flavodoxin (3 mM) in the oxidized, the semiquinone, and the hydroquinone states in 150 mM Tris-HCl, pH 8.2. (A) Oxidized state (line broadening 1 Hz). (B) Mixture of about 30% oxidized and about 70% semiquinone form (line broadening 1 Hz). (C) Semiquinone form (line broadening 3 Hz). (D) Mixture of about 21% semiquinone and about 80'% hydroquinone form (line broadening 3 Hz). (E) Hydroquinone form (line broadening 1 Hz).
ilP NMR on M. elsdenii Flavodoxin. The 31P NMR spectrum of M. elsdenii flavodoxin is shown in Figure 3A. The spectrum exhibits only one phosphorus resonance at -4.80 ppm due to bound FMN in contrast to other flavodoxins containing more .than one phosphate group (Edmondson & James, 1979). Free FMN in the dianionic form shows a chemical shift of-4.70 ppm. The chemical shift of protein-bound FMN suggests therefore that the phosphate group of FMN is bound to the protein in the dianionic form. The small downfield shift of protein-bound FMN as compared to that of free FMN is probably due to steric effects or strain. Such a downfield shift due to strain has been observed in alkaline phosphatase (Bock & Sheard, 1975).
The 31P chemical shift of M. elsdenii flavodoxin is independent of the pH in the range 6.0-9.2. Below pH 6.0 a small upfield shift is observed which amounts to 0.3 ppm at pH 5.5. At pH values below 5.5, the solubility of the flavodoxin decreases [pi is about 4 (Gast et al., 1976)] preventing studies at lower pH values. The results indicate that the phosphate group of protein-bound FMN is deeply buried in the apo-enzyme and unaccessible to bulk solvent.
11
"P NMR OF M. E L S D E N I I F L A V O D O X I N
The proton-coupled 31P NMR spectrum of free FMN shows a triplet due to coupling of the 5'-CH2 group with the 5'-phosphorus atom (data not shown). The vicinal coupling constant is about 7.2 Hz. The proton-coupled spectrum of M. elsdenii flavodoxin does not show a spitting of the phosphorus resonance line, but the width of the resonance line is about 5.5 Hz. The proton-decoupled spectrum, on the other hand, exhibits a line width of 2.3 ± 0.2 Hz. These observations can be analyzed by a Karplus-like relation (Cozzone & Jardetzky, 1976) in terms of the structure of the ribityl phosphate bonding and indicate a gauche-gauche conformation. Similar results were reported by Favaudon et al. (1980) for flavodoxin from D. vulgaris and D. gigas. The latter results and those presented here are in agreement with three-dimensional data on Clostridium MP flavodoxin (Burnett et al., 1974).
The influence of reduction of an anaerobic solution of M. elsdenii flavodoxin on the 31P NMR spectrum is shown in Figure 3. During the addition of the first electron to oxidized flavodoxin, yielding quantitatively the flavosemiquinone (Mayhew, 1978), the resonance line due to oxidized flavodoxin is decreased and broadened. In the semiquinone form only a broad peak (5 ~ -4.9 ppm) is observed. Further reduction of the flavosemiquinone yielding finally the flavohydroquinone does not alter the NMR spectrum until almost full reduction is achieved. The 31P NMR spectrum of the flavohydroquinone exhibits a sharp resonance line at -4.9 ppm. The observed spectral changes are fully reversible upon stepwise reoxidation of the solution of reduced flavodoxin. The broadening of the resonance line of the phosphate group of bound FMN must be due to the flavin radical since other effects, such as drastic conformational changes of the protein, can be excluded by three-dimensional data obtained for the related flavodoxin from Clostridium MP (Andersen et al., 1972). The small difference between the 31P chemical shifts of the three redox states of flavodoxin indicates that the binding interaction between the phosphate group of FMN and the apoprotein is not perturbed by redox reactions of the isoalloxazine moiety. This conclusion is in agreement with X-ray data on Clostridium MP flavodoxin (Andersen et al., 1972). 31P NMR data similar to that described above were found for D. vulgaris and D. gigas flavodoxin by Favaudon et al. (1980).
Analysis of the line width observed during oxidation-reduction experiments yields information with respect to the rate of electron transfer between flavodoxin molecules of different redox state. By analogy to chemical exchange reactions (Dwek, 1973; McLaughlin & Leigh, 1973), the perturbation of X, and T2 of the phosphorus resonance can be correlated with the rate of transfer of one electron between two molecules of flavodoxins. Reduction of the oxidized flavodoxin to the semiquinone state does not affect the resonance line of the oxidized flavodoxin, but it is superimposed on the broad line of the semiquinone form (Figure 3A-C). Such a situation is characterized by a slow exchange reaction. The electron transfer reaction can be described by the equilibrium reaction
Fl„ml + Fl,- ^ Fl,- + Fl„„Tl CD ' l (ox) '2(ox)
where Fl1(ox) and Fl2- are two flavodoxin molecules in the oxidized and semiquinone (F1-) state, respectively. The electron-transfer reaction is a second-order process. The experimental data were analyzed according to McLaughlin & Leigh (1973). Since no broadening of the resonance line of the oxidized flavodoxin, is observed, i.e., T2 is not affected, by reduction to the semiquinone form, the limit of the lifetime T is calculated to be »0.5 s .James et al. (1973) calculated from 'H NMR data a value for T »0.02 s . This latter value is muchsmal 1 erthan our value, but the
VOL. 21 , N O . 2 , 1982 411
experimental data of James et al. (1973) did not allow a more accurate calculation of kacll.
In going from the semiquinone to the hydroquinone form of flavodoxin, no superimposed lines are observed in the 31P NMR spectrum. Even in the presence of only a very small concentration of flavosemiquinone in the solution studied only a broad resonance line is observed. In fact the spectrum shown in Figure 3D was obtained by admission of a small volume of air to the solution of the flavodoxin hydroquinone, yielding a small concentration of flavosemiquinone (about 20 %). The fact that no sharp line could be observed indicates that we are dealing here with a fast-exchange reaction. Analysis of these data according to McLaughlin & Leigh (1973) yields a lower limit of the lifetime x of «0.01 s . Thesmall difference between the "P chemical shifts of the oxidized and semiquinone form of flavodoxin, on the one hand, and that between the semiquinone and hydroquinone form, on the other hand, does not allow a more accurate analysis of the rates of electron transfer. Nevertheless, the experimental results show clearly that under our experimental conditions a large difference exists in the rate of electron transfer between molecules of oxidized and semiquinone flavodoxin and that between molecules of semiquinone and hydroquinone flavodoxin. This is the first direct proof that fast electron transfer occurs under equilibrium conditions between semiquinone and hydroquinone flavodoxin molecules. James et al. (1973) also reported that the lifetime forthe transfer betweensemiquinone and hydroquinone molecules is the same as that between the oxidized and semiquinone forms, i.e. »0.02 s .This is in contradiction with our results and is caused by the severe limitations of the analysis of the 'H NMR data.
The kinetics of the two step one-electron reduction of M. elsdenii flavodoxin by dithionite was investigated by Mayhew & Massey (1973). These authors found that the rate of reduction of the semiquinone to the hydroquinone form is at least 2 orders of magnitude larger than that of the reduction of the oxidized to the semiquinone state. On the other hand, the potential of the semiquinone-hydroquinone redox couple is more negative than that of the quinone-semiquinone redox couple, suggesting that the latter redox reaction should be more favored energetically than the former one. The kinetic data of Mayhew & Massey (1973) and our own data strongly indicate that the activation energy for the reduction of the semiquinone and the electron exchange between semiquinone and hydroquinone is much smaller than that for the quinone-semiquinone couple. These observations may be explained by a protein conformational change occurring during the quinone-semiquinone transition and not during the semiquinone-hydroquinone transition, resulting in a higher transition energy for the first transition. The X-ray results of Andersen et al. (1972) support this suggestion. The formation or breakage of specific interactions between the prosthetic group and the apoprotein may also cause small local conformational (configurational?) changes of some group(s). The problem is therefore of kinetical rather than thermodynamic origin. Our electron-exchange studies now also offer an explanation for the fact that the comproportionation reaction, i.e., Fl0» + Flred i=± 2F1-, in flavodoxin is not favored (Mayhew & Massey, 1973). From a kinetic point of view, taking into account our results, the electron-exchange reaction between quinone and hydroquinone must be orders of magnitude smaller than that calculated for the quinone-semiquinone transition.
The biological function of flavodoxins is to transfer one electron at a time to other redox proteins (e.g., hydrogenase)
12
4 1 2 B I O C H E M I S T R Y
shuttling thereby between the semiquinone and hydroquinone state (Mayhew & Ludwig, 1975, and references therein). As can be calculated from recently published results of Van Dijk & Veeger (1981), the turnover number of M. elsdenii flavodoxin in a reaction with hydrogenase can be as high as 105
s"1. The data of Mayhew & Massey (1973) and our own data allow some rationalization of these biologically important reactions.
The strong broadening of the phosphorus resonance in the NMR spectrum of flavodoxin in the semiquinone form (Figure 3C) allows us to calculate the distance between the isoall-oxazine ring and the phosphorus atom of bound FMN. Since the electron-nuclear hyperfine coupling on the phosphorus atom can be neglected (no spin density of the radical on this nucleus), we have to consider only the dipolar part of the original Solomon-Bloembergen (Solomon, 1955; Bloembergen, 1957) equation.
The rotational correlation time of flavodoxin M. elsdenii was calculated according to the Stokes-Einstein equation, taking into account the solvation of the molecule (Tanford, 1961). The rotational correlation time was calculated to be 5 x 10"' s at 26 °C (temperature of NMR experiments). Analytical ultracentrifugation experiments were carried out in order to check if the concentrations (3 mM) used in the NMR experiments lead to association of flavodoxin molecules. No aggregation could be detected, so the calculated TC value is a good approximation. The 7"2p (paramagnetic spin-spin relaxation) value was calculated from the line width of spectra of the flavodoxin. From Figure 3B a line width of 43 Hz is calculated. The line width in the spectrum of oxidized flavodoxin yields the diamagnetic contribution to the line width, i.e., field inhomogeneity, 'H-31P dipole-dipole interaction, and chemical shift anisotropy. In the semiquinone form these contributions are approximately the same. This procedure yields a line width of about 37 Hz (taking into account the exponential line broadening) for the paramagnetic contribution, from which a Tlp [=(irAi/1/2)~
1] of 8.6 X 10~3 s is calculated. With the aid of these values the distance between the isoall-oxazine radical and the phosphorus nucleus of FMN was calculated to be 8.8 Â (0.88 nm). The highest spin density in flavin radicals is located in the pyrazine subnucleus of the isoalloxazine molecule (Müller et al., 1971). If we assume for convenience that the average effective location of the free electron is at about 1-Â distanced from the N(10) atom of flavin, i.e., centered in the pyrazine subnucleus, then the calculated distance between the N(10) atom and the phosphorus nucleus of FMN is about 7.8 Â. This is in good agreement with the distance between the two atoms (8.5 Â) as obtained by X-ray studies for the Clostridium MP flavodoxin (Burnett et al., 1974). The rather good agreement between the two independent methods indicates that the side chain of FMN is strongly bound, with little or no internal freedom. If the side chain had possessed some internal mobility, we would have been forced to take into account an additional TC value for the internal mobility which would influence the calculated value for the distance considerably.
According to the three-dimensional structure of Clostridium MP flavodoxin (Burnett et al., 1974), no charged groups are located in the immediate neighborhood of the phosphate group of FMN. There are, however, various polar amino acid residues in the vicinity of the phosphate group, i.e., four hy-droxyamino acid residues and five backbone NH protons. In addition, one molecule of water is possibly also in the vicinity of the phosphate group. For investigation of the environment of the phosphate group, a comparative relaxation study be-
M O O N E N A N D MULLER
FIGURE 4: Partially relaxed 31P NMR spectra at 40.5 MHz of M. elsdenii flavodoxin (3 mM) in 150 mM Tris-HCl, pH 8.2, obtained by the 180°-T-90° pulse sequence and Fourier transformation. The labile protons were exchanged prior to the experiments (cf. Materials and Methods).
Table II: Calculated T, and r, Relaxation Times of Oxidized M. elsdenii Flavodoxin
protein preparation r,°(s) r2°(s) native protein protein reconstituted
in deuterium oxide c
0.57 ± 0.06 1.17 ± 0.12
0.14 ± 0.02 0.21 + 0.03
° Determined by the inversion-recovery technique (accuracy estimated). b Determined from the line width (accuracy estimated). c Cf. Materials and Methods.
tween native flavodoxin and flavodoxin reconstituted in deuterium oxide (cf. Materials and Methods) was performed. T, and T2 measurements should make it possible to elucidate the influence of labile protons in the vicinity of the phosphate group on the relaxation times. Figure 4 shows spin-lattice relaxation experiments performed with flavodoxin reconstituted in deuterium oxide. Fitting of the data by one exponential curve yields a 7", value of 1.17 s. In Table II the calculated 7, and T2 values are presented. From the 7"2 values of Table II it is obvious that the line width of the phosphorus resonance is different in the two preparations studied. The chemical and the physical properties of the two preparations are identical, so it must be concluded that the difference in line width must originate from the influence of labile protons on the T2 relaxation.
For studies on a possible back exchange of deuterons in reconstituted protein, a sample was dissolved in a mixture of H 20/ 2H 20 (9:1 v/v) in 150 mM Tris, pH 8.0, and the line width followed with time. It was found that up to 20 h the line width remained constant, i.e., no back exchange occurred. A similar experiment (10 h) with the hydroquinone form of flavodoxin yielded the same result. This means that the exchange reaction, if it occurs at all, must be very slow. Also these results indicate that the phosphate group of FMN interacts very strongly with the apoprotein and that the exchange of FMN molecules in flavodoxin is a very slow process.
- 13 -
"f NMR OF M. ELSDENII FLAVODOXIN
For oxidized, native flavodoxin the spin-lattice relaxation can be described by
1 • ' ' * ( 2 ) 'l('H,exch) l(rcst)
where r1(iHeiIch) is the relaxation by labile protons in native flavodoxin, 7"1(iHll0n„ch) is the relaxation by nonexchangeable protons in flavodoxin reconstituted in deuterium oxide, and 7"i(rest) >s t n e relaxation by all other mechanisms. For the reconstituted flavodoxin the relaxation by labile protons is eliminated whereas the relaxation by deuterons can be neglected because of the low gyromagnetic ratio. The relaxation by labile protons can therefore be expressed as
1 = ! ! O) T , T T 1 l('H,exch) l 1 (native flavod) I 1 (reconst flavod)
Using the values given in Table II and eq 3 yields a value of 1.11 ± 0.3 s for the relaxation by labile protons (7"1(iHcjCh). With the aid of eq 4 (Abragam, 1961) the calculated value
l _ /"ft 2rHy\ x TH'H,tich) \ lOrpH6 J
U 3TC 67^ I
1 + (ü>H - ü>p) V 1 + ü>p2Tc2 1 + («H + U)p)2Tc
2 J
(4) can be correlated with the number of labile protons and their distance from the phosphorus atom of FMN. n is the number of protons separated by a distance rfli from the phosphorus nucleus, 7H and yP are the gyromagnetic ratios of the proton and phosphorus nucleus, respectively, uH and <iif are the corresponding Larmor precession frequencies, TC is the rotational correlation time, and h is Plank's constant. It should be mentioned that eq 4 does not allow us to determine the number of protons accurately, but only an estimate can be made. Moreover the distance of the protons from the phosphorus atom is assumed to be the same for all labile protons, but in reality this distance may be different for different labile protons. This is important to notice since because of the sixth power dependence, the distance between protons and the phosphorus atom is of great influence on TV For the distance rPH a value of 2.75 Â was assumed. With these values the number of labile protons was calculated to be five (eq 4). Considering the approximate value of rPH the calculated value of five protons is in fair agreement with crystallographic data (Burnett et al., 1974) on the flavodoxin from Clostridium MP. This result confirms the conclusion drawn above with respect to the strong interaction of the phosphate group of FMN with the apoflavodoxin from M. elsdenii.
With the results described above we can finally test if the decrease of the line width observed in reconstituted flavodoxin as compared to that in native flavodoxin is in agreement with the T\ data (Table II). For the analysis of the T2 experiment, equations analogous to eq 2 and 3 can be written. In this way we obtain a value of 0.42 ± 0.15 s for the spin-spin relaxation by labile protons. For analysis of this value, the following equation is applied (Abragam, 1961):
V 20rP„' / [ 4T„ +
3rc 6TC
1 + (UH - 0)p)2Tc2
6T„
1 + »p2Tc2 1 + a>H
2Tc2 1 + (Û>H + <»p)2V
(5)
The symbols in this equation are the same as in eq 4. When « = 5 and rPH = 2.75 Â, a theoretical value of 0.6 s for the
VOL. 21 , NO. 2, 1982 413
spin-spin relaxation by labile protons is calculated. This value is in fair agreement with the experimental value of 0.42 s This means, therefore, that the results of the analysis of T2 support the data of the analysis of Tt. From this it can be concluded that labile protons interact via hydrogen bonding with the phosphate group of protein-bound FMN, but the results do not allow a precise determination of the number of protons involved in the hydrogen bonding and their distance to the phosphorus nucleus. The agreement between our results and X-ray data also indicates that the ribityl phosphate group of FMN is strongly bound to the M. elsdenii apoflavodoxin and possesses no measurable internal mobility.
Finally it is hoped that the 3IP NMR technique, in combination with other physical methods, will provide a deeper insight into the structure-function relationship of flavoproteins.
Acknowledgments We thank Dr. G. Voordouw for assistance in the analytical
ultracentrifugation experiments, Dr. C. van Dijk for invaluable discussions, J. C. Toppenberg-Fang for the typing of the manuscript, W. A. M. van der Berg for the isolation of the flavodoxin, M. M. Bouwmans for the preparation of the figures, and Dr. G. Searle for linguistic advice.
References
Abragam, A. (1961) The Principles of Nuclear Magnetism, pp 289-298, Clarendon Press, Oxford.
Andersen, R. D., Apgar, P. A., Burnett, R. M., Darling, G. D., Lequesne, M. E., Mayhew, S. G., & Ludwig, M. L. (1972) Proc. Natl. Acad. Sei. U.S.A. 69, 3189-3191.
Bloembergen, N. (1957) J. Chem. Phys. 27, 572-573, 596. Bock, J. L., & Sheard, B. (1975) Biochem. Biophys. Res.
Commun. 66, 24-30. Burnett, R. M., Darling, G. D., Kendall, D. S., LeQuesne, M.
E., Mayhew, S. G., Smith, W. W„ & Ludwig, M. L. (1974) J. Biol. Chem. 249, 4383-4392.
Cozzone, P. J., & Jardetzky, O. (1976) Biochemistry 15, 4853-4865.
Dwek, R. A. (1973) Nuclear Magnetic Resonance in Biochemistry, pp 174-212, Clarendon Press, Oxford.
Edmondson, D. E„ & James, T. L. (1979) Proc. Natl. Acad. Sei. U.S.A. 76, 3786-3789.
Evans, F. E., & Kaplan, N. O. (1979) FEBS Lett. 105, 11-14. Favaudon, V., Le Gall, J., & Lhoste, J.-M. (1980) in Flavins
and Flavoproteins (Yagi, K., & Yamano, T., Eds.) pp 373-386, Japan Scientific Societies Press, Tokyo.
Gast, R., & Müller, F. (1978) Helv. Chim. Acta 61, 1353-1363.
Gast, R., Valk, B. E., Müller, F., Mayhew, S. G., & Veeger, C. (1976) Biochim. Biophys. Acta 446, 463-471.
Gorenstein, D. G. (1975) J. Am. Chem. Soc. 97, 898-900. Gorenstein, D. G., Wyswicz, A. M., & Bode, M. (1976) J.
Am. Chem. Soc. 98, 2308-2314. James, T. L., Ludwig, M. L„ & Cohn, M. (1973) Proc. Natl.
Acad. Sei. U.S.A. 70, 3292-3295. Levy, G. C , & Peat, I. R. (1975) J. Magn. Reson. 18,
500-521. Mayhew, S. G. (1971) in Flavins and Flavoproteins (Kamin,
H., Ed.) pp 185-209, University Park Press, Baltimore, MD. Mayhew, S. G. (1978) Eur. J. Biochem. 85, 535-547. Mayhew, S. G., & Massey, V. (1969) J. Biol. Chem. 244,
794-892. Mayhew, S. G., & Massey, V. (1973) Biochim. Biophys. Acta
315, 181-190. Mayhew, S. G., & Ludwig, M. L. (1975) Enzymes, 3rd Ed.
12, 57-118.
14
414 Biochemistry 1982, 21, 414-419
Mayhew, S. G., & Strating, M. J. J. (1975) Eur. J. Biochem. 59, 539-544.
McLaughlin, M. A., & Leigh, J. (1973) J. Magn. Resort. 19, 296-304.
Müller, F. (1972) Z. Naturforsch. B: Anorg. Chem., Org. Chem., Biochem., Biophys., Biol. 27B, 1023-1026.
Müller, F., Hemmerich, P., & Ehrenberg, A. (1971) in Flavins and Flavoproteins (Kamin, H., Ed.) pp 107-122, University Park Press, Baltimore, MD.
Müller, F., Mayhew, S. G., & Massey, V. (1973) Biochemistry 12, 4654-4662.
Nanda, R. K., Ribeiro, A., Jardetzky, T. S., & Jardetzky, O. (1980) J. Magn. Reson. 39, 119-125.
Sarma, R. H., Dannies, P., & Kaplan, N. O. (1968) Biochemistry 7, 4359-4367.
Sass, M., & Ziessow, D. (1977) J. Magn. Reson. 25, 263-276. Scola-Nagelschneider, G., & Hemmerich, P. (1976) Eur. J.
Biochem. 66, 567-577. Smith, W. W., Burnett, R. M., Darling, G. D„ & Ludwig,
M. L. (1977) J. Mol Biol 117, 195-225. Solomon, I. (1955) Phys. Rev. 99, 559-565. Tanford, C. (1961) Physical Chemistry of Macromolecules,
pp 317-456, Wiley, New York. Van Dijk, C , & Veeger, C. (1981) Eur. J. Biochem. 114,
209-219. Void, R. L., Waugh, J. S., Klein, M. P., & Phelps, D. E.
(1968) J. Chem. Phys. 48, 3831-3832. Wassink, J. H., & Mayhew, S. G. (1975) Anal. Biochem. 68,
609-616.
15
Chap te r 3
Volume 149, number 1 FEBS LETTERS November 1982
A photo-CIDNP study of the active sites of Megasphaera elsdenii and Clostridium MP flavodoxins
Chrit T .W. Moonen, Peter J . H o r e + * , Franz Müller, Robert Kaptein+ and Stephen G. Mayhew+
Department of Biochemistry, Agricultural University, De Dreijen 11, 6703 BC Wageningen, * Department of Physical Chemistry, University of Groningen, Nijenborgh 16, 9747 AG Groningen, The Netherlands and
^Department of Biochemistry, University College, Belfield, Dublin 4, Ireland
Received 30 September 1982
Megasphaera elsdenii and Clostridium MP flavodoxins have been investigated by photo-CIDNP techniques. Using time-resolved spectroscopy and external dyes carrying different charges it was possible to assign unambiguously the resonance lines in the NMR-spectra to tyrosine, tryptophan and methionine residues in the two proteins. The results show that Trp-91 in M.elsdenii and Trp-90 in CI.MP flavodoxin are strongly immobilized and placed directly above the benzene subnucleus of the prosthetic group. The
data further indicate that the active sites of the two flavodoxins are extremely similar.
Flavodoxin Megasphaera elsdenii Clostridium MP Photo-CIDNP NMR
1. INTRODUCTION
The flavodoxins are a class of small proteins (M, 15000-23000) which contain the prosthetic group riboflavin 5'-phosphate (FMN) and function as electron carriers in biological reactions [1]. Chemical modification [2,3], X-ray crystallography [4] and fluorescence quenching [3] studies have shown that a number of aromatic amino acid residues are located in the neighbourhood of the flavin binding site arid that these play an important role in the interaction between prosthetic group and apo-enzyme. Here, we report a photo-CIDNP investigation of these aromatic residues in two closely related flavodoxins from the bacteria Megasphaera elsdenii and Clostridium MP.
• Present address: Physical Chemistry Laboratory, South Parks Road, Oxford, England
Abbreviations: photo-CIDNP, photochemically induced dynamic nuclear polarization; NMR, nuclear magnetic resonance; flavin I, 3-JV-carboxymethyllumiflavin; flavin II, S-N-ethylaminolumiflavin; FID, free induction decay
The photo-CIDNP method [5,6] is based on the generation of nuclear spin polarization in a reversible reaction between the protein and a photo-excited dye. When accessible to the dye, the side chains of histidine, tryptophan and tyrosine residues can be polarized resulting in selective enhancements in the 'H NMR spectrum of the protein. A light-minus-dark difference technique leads to a photo-CIDNP spectrum containing only resonances from polarized nuclei [7].
An earlier investigation of the M.elsdenii and Clostridium MP flavodoxins showed that several tyrosine and tryptophan residues are polarizable [8]. Here, we extend this work to compare the active sites of the two flavodoxins, exploiting the dependence of CIDNP intensities on the timing of the experiment and on the nature of the dye to arrive at unambiguous NMR assignments as well as indications of the mobilities and exposure of the polarizable residues. Our results demonstrate that the active sites of the two proteins are extremely similar. As the spectra presented here are qualitatively much better than published for M. elsdenii flavodoxin in [8] chemical shifts could be determined much more accurately. This is the reason for
Published by Elsevier Biomedical Press 001457593/82/0000—0000/S2.75 © Federation of European Biochemical Societies 141
- 16
Volume 149, number 1 FEBS LETTERS November 1982
the minor differences between chemical shifts reported in [8] and those presented here.
2. MATERIALS AND METHODS
Megasphaera elsdenii and Clostridium MP flavo-doxins were isolpted and purified as in [9]. NMR samples consisted of 1 mM protein with 100 mM potassium phosphate buffer in D20 at pH 8.0. 3-7V-Carboxymethyllumiflavin (flavin I) and 3-N-ethylaminolumiflavin (flavin II) were the dyes used for CIDNP generation. All spectra were recorded at 360 MHz on a Bruker HX-360 NMR spectrometer [5,7]. Ten free induction decays were ac
cumulated for each spectrum. Chemical shifts are quoted with respect to trimethylsilylproprionate (TSP).
3. RESULTS AND DISCUSSION
Several photo-CIDNP spectra of M.elsdenii flavodoxin, obtained with the positively charged dye (flavin II), are presented in fig. 1. Two tyrosines and two tryptophans are polarized, the assignments coming directly from the characteristic CIDNP phases [5] and the fact that the molecule contains no histidine [10]. The distinction between resonances from the two Trp residues was achieved
)k?{%/AlHL /r^sHvy^A/W
LIGHT PERIOD
0.1 s
0.2 s
DELAY
0.05 s
0.2 s
0.6 s O./f s
1.0 s 1.0 s
1.0 s 1.5 s
PPM (s)
Fig. 1. Photo-CIDNP spectra of M.elsdenii flavodoxin with 0.5 mM flavin II as a function of the light period and delay time, as indicated: Trp I (O); Trp II (A); Tyr I (o); Tyr II (A). The HDO resonance at 4.8 ppm (arrow) has been
omitted.
142
Volume 149, number
- 17 -
FEBS LETTERS November 1982
by comparison of spectra (not shown) generated with the negatively charged dye (flavin I). With this dye, the lines of Trp I (o,#) decreased by - 5 0% relative to those of Trp II (A) . These assignments are supported and completed by considering cross relaxation effects, as follows.
Of the Trp sidechain protons it is known that only H-2, H-4, H-6 and H-ß are appreciably polarized in photoreactions with flavins. Any substantial CIDNP effects observed for H-5, H-7 and H-a arise by dipolar cross relaxation with directly polarized protons and are thus said to be cross- (or indirectly)-polarized [11]. The two types of polarization are easily distinguished by their time dependence [11]. With a short light flash and short delay before FID acquisition (upper spectrum in fig. 1) the indirect signals are very weak while with longer light and delay times they become stronger relative to their directly enhanced counterparts. For example, the resonances at 4.88, 6.41 and 6.80 ppm (fig. 1) are strongly cross-polarized whereas those at 5.57, 7.09 and 7.61 ppm are directly polarized. Such arguments combined with the line multiplicities enable us to assign all but the very weakest resonances visible in fig. 1 as summarized in table 1.
Further information can be extracted from fig. 1. First we note that cross relaxation is much less extensive in Trp II than in Trp I: indeed the H-a, H-5 and H-7 resonances of the former were too weak to be identified with certainty. This, together with the relative narrowness of its resonances, suggests that of the two, Trp II is considerably more mobile. Similarly, the two Tyr residues which exhibit relatively long relaxation times must also enjoy a fair degree of mobility.
The cross relaxation effects in Trp I are rather pronounced especially for H-a and H-2 both of which receive polarization from the emissively enhanced ß protons. Indeed the effect is sufficiently great for H-2 to cause it to change phase to emission with increasing light and delay periods. This observation, although not unexpected theoretically [11], has not been seen before and implies that Trp I is strongly immobilized in M.elsdenii flavo-doxin. To obtain some estimate of the dihedral angles [12], xi and Xi< which also influence cross polarization between the ß protons and H-a and H-2, respectively, we simulated the evolution of the CIDNP intensities expected for a tumbling
Table 1
Assignments of resonances observed in the photo-CIDNP spectra of Clostridium MP and M.elsdenii
flavodoxins
Trp I H-/3A
H-fo H-a H-2 H-4 H-5 H-6 H-7
Trp II
H-A H-/ÎB H-2 H-4 H-5 H-6 H-7
Tyr I H-3, H-5
Tyr II H-3, H-5
M.elsdenii Trp-91
3.38 3.38 4.88 7.09 7.61 6.80 5.57 6.41
Trp-7 3.45 3.31 7.31 7.72
? 7.27
?
Tyr-6 or Tyr-89 6.90
Tyr-6 or Tyr-89 6.88
Clostridium MP
Tyr
Trp-90
3.34 3.34 4.98 7.07 7.55 6.80 5.62 6.35
Trp-6 3.52 3.11 7.18 7.55 6.99 7.12 7.43
-5 or Tyr-88 6.65
Tyr-5 or Tyr-88 6.58
tryptophan residue with a rotational correlation time of 5 ns [13]. Of the various pairs of x\, Xi for which the calculation was performed, our experimental results agree most closely with x\ = 60° and xi = ±90°. The value for x\ is consistent with the failure to detect coupling for the H-a- resonance (i .e., J < 6 Hz) indicating that H-a is approximately gauche with respect to both ß protons.
Fig. 2 shows photo-CIDNP spectra of Clostridium MP flavodoxin recorded with flavin II (2A) and flavin 1 (2B,2C) as external dyes. Most of the resonances were assigned (table 1) by the above methods with resolution enhancement [14] to determine the multiplet patterns. The unusual chemical shifts and the cross-polarization effects for Trp I here are very similar to M.elsdenii; however, it is clear that the charge of the dye plays a more important role in the Clostridium MP pro-
143
Volume 149, number 1 FEBS LETTERS November 1982
LIGHT PERIOD
0.6 s
0.1 s
0.6 s
PPM [()
Fig. 2. Photo-CIDNP spectra of Clostridium MP flavodoxin with 0.5 mM flavin H (A) and 0.5 mM flavin I (B,C). A delay of 0.05 s was used with light periods as indicated. Notation as in fig. 1.
tein. With the negative flavin the enhancements of Trp I are much weaker while Trp II is completely absent.
We now consider the assignment of the observed resonances to particular amino acid residues in the primary sequences of the two proteins. M.elsdenii flavodoxin [10] contains four tryptophans at positions 7, 91, 96 and 100 whereas Clostridium MP [15] has three at 6, 90 and 95. The last three are invariant in the two proteins.
The solution to this problem comes from the dramatic upfield shifts (1.6 ppm for H-6 and 1.1 ppm for H-7) from the random coil positions experienced by some of the Trp I protons. Almost certainly due to ring current effects, such shifts require the affected protons to be 0.3-0.4 nm directly above the plane of a neighbouring aroma
tic ring [16]. Inspection of the crystal structure of Clostridium MP flavodoxin [4] shows Trp-90 to be the most obvious candidate. This residue is completely exposed to the solvent on one side and very close to the ring system of the protein-bound flavin on the other. The observed chemical shift of the individual protons of Trp I are in excellent qualitative agreement with this conclusion. Moreover reduction of the flavodoxin to its paramagnetic semiquinone form broadens the resonance of Trp I [17] providing independent support for the correctness of this assignment.
The similarity in behaviour and chemical shifts exhibited by Trp II in the two flavodoxins points to an invariant residue. The X-ray data for Clostridium MP flavodoxin indicate that Trp-6 is partially exposed but that Trp-95 is buried in the
144
Volume 149, number 1
- 19 -
FEBS LETTERS November 1982
interior of the protein and therefore probably inaccessible to either flavin dye. We therefore tentatively assign Trp II to Trp-6 in Clostridium MP and Trp-7 in M.elsdenii flavodoxin. This conclusion is supported by the observed accessibility of Trp II to the oppositely charged flavin dyes. In Clostridium MP flavodoxin, the carboxylate group of Glu-65 is close to Trp-6 and would facilitate the approach of flavin II but hinder that of flavin I. In M.elsdenii flavodoxin, where we do not observe such a strong dependence on the charge of the dye, this Glu is replaced by a valine.
A further interesting feature of fig. 1 is the sharp, indirectly polarized line at 2.15 ppm. A corresponding (although somewhat broader) resonance from Clostridium MP flavodoxin was found at 2.02 ppm. Showing a dependence on the charge of the dye similar to Trp II, it appears to receive its polarization from that residue, while its linewidth and position strongly suggest a methionine f-methyl group. We assign this resonance tentatively to Met-56 in Clostridium MP flavodoxin (Met-57 in M. elsdenii) which has e-protons <0.3nm from Trp-6(7) in the crystal structure.
Turning to the tyrosine residues (positions 6 and 89 in M.elsdenii flavodoxin and 5 and 88 for Clostridium MP), the two directly polarized emissive doublets in the spectra of fig. 1 must arise from the H-3,5 protons of the two tyrosines, although one cannot say which is which. Interestingly these tyrosines in Clostridium MP flavodoxins are only weakly polarized by flavin II (at 6.65 and 6.58 ppm) and hardly at all by flavin I. This observation could be interpreted as evidence either for a lower accessibility of these residues or for some hydrogen bonding interaction of their hydroxyl groups in the Clostridium MP protein.
Finally, a number of as yet unidentified lines can be seen in the photo-CIDNP spectra of the two proteins. In fig. 1 weak absorptive resonances are visible at 7.56, 7.63, 7.46, 7.21, 7.18 ppm and in fig. 2 at 7.87 and 5.54 ppm. Some of these may arise from a third, weakly enhanced tryptophan, from the H-5 or H-7 protons of Trp II in the case of M.elsdenii flavodoxin or from other residues cross polarized from Trp I or Trp II.
The most important conclusions to come out of this study involve the active site residue Trp 89 (90). It shows almost identical behaviour in the two flavodoxins and enjoys little internal mobility. In
M.elsdenii flavodoxin the isoalloxazine ring is also immobilized [18]. The question now arises as to what role this rigid isoalloxazine-tryptopohan complex plays in the electron transfer function of these flavoproteins.
ACKNOWLEDGEMENTS
We thank Miss C M . Verstege for typing the manuscript, Mr M.M. Bouwmans for the preparation of the figures, Mr W.A.M. van den Berg for the isolation of M.elsdenii flavodoxin and Mr K. Dijkstra for excellent technical assistance. This work was supported in part by the Netherlands Foundation for Chemical Research (SON) with financial aid from the Netherlands Organization for the Advancement of Pure Research (ZWO). P.J.H, is grateful to the Royal Society of London for the award of a research fellowship.
REFERENCES
[1] Mayhew, S.G. and Ludwig, M.L. (1975) in: The Enzymes, 3rd edn, vol. 12 (Boyer, P.D. ed) pp. 57-118, Academic Press, New York.
[2] McCormick, D.B. (1970) Experimentia 26, 243-244.
[3] Mayhew, S.G. (1971) Biochim. Biophys. Acta 235, 289-301.
[4J Burnett, R.M., Darling, G.D., Kendall, D.S., LeQuesne, M.E., Mayhew, S.G., Smith, W.W. and Ludwig, M.L. (1974) J. Biol. Chem. 249, 4383-4392.
[5] Kaptein, R. (1978) in: NMR Spectroscopy in Molecular Biology (Williams, G.D. ed) vol. 5, pp. 211-229, D. Reidel, Dordrecht.
[6] Kaptein, R. (1982) in: Biological Magnetic Resonance (Berliner, L.J. and Reuben, J. eds) vol. 4, pp. 145-191, Plenum, New York.
[7] Kaptein, R., Dijkstra, K., Müller, F., Van Schagen, CG. and Visser, A.J.W.G. (1978) J. Magn. Res. 31, 171-176.
[8] van Schagen, CG., Müller, F. and Kaptein, R. (1982) Biochemistry 21, 402-407.
[9] Mayhew, S.G. and Massey, V. (1969) J. Biol. Chem. 244, 794-802.
[10] Tanaka, M., Haniu, M., Yasunobi, K.T., Mayhew, S.G. and Massey, V., J. Biol. Chem. (1973) 248, 4354-4366; (1974) 249, 4397.
[11] Höre, P.J., Egmond, M.R., Edzes, H.T. and Kaptein, R. (1982) J. Magn. Res. in press.
[12] IUPAC-IUB Commission on Biochemical Nomenclature, Biochemistry (1970) 9, 3471-3479.
145
- 20 -
Volume 149, number 1 FEBS LETTERS November 1982
[13] Moonen, C.T.W, and Müller, F. (1982) Biochemis- [16] Perkins, S.J. and Wüthrich, K. (1979) Biochim. try 21, 408-414. Biophys. Acta 576, 409-423.
[14] Ferrige, A.G. and Lindon, J.C. (1978) J. Magn. [17] van Schagen, CG. and Müller, F. (1981) FEBS Res. 31, 337-340. Lett. 136, 75-79.
[15] Tanaka, M., Haniu, M., Yasunobu, K.T. and [18] Moonen, C.T.W, and Müller, F. (1983) submitted. Mayhew, S.G. (1974) J. Biol. Chem. 249, 4393-4396.
146
21
Eur. J. Biochem. 133, 463-470 (1983) ©FEBS1983
Chapter 4
On the Mobility of Riboflavin 5'-Phosphate in Megasphaera elsdenii Flavodoxin as Studied by 13C-Nuclear-Magnetic-Resonance Relaxation
Chrit T. W. MOONEN and Franz MÜLLER
Department of Biochemistry, Agricultural University, Wageningen
(Received December 20, 1982/March 22, 1983) - EJB 6352
The mobility of the isoalloxazine ring of the prosthetic group of Megasphaera elsdenii flavodoxin was investigated by a 13C relaxation study of the non-protonated ring atoms 2, 4, 4a and 10a. In this study a selectively enriched (> 90% 13C) prosthetic group was bound to the apoprotein. T\ and T2 values were determined at two magnetic field strengths, i.e. 8.46 T (90.5 MHz) and 5.88 T (62.8 MHz). Values of nuclear Overhauser effects (NOE) were determined at 5.88 T. It is shown that both the dipole-dipole interaction and the chemical shift anisotropy are important relaxation sources for all the carbon atoms investigated. The results are in agreement with a spectral density function of the isoalloxazine ring in which only the overall reorientational motion of the protein is accounted for. From this it is concluded that the isoalloxazine ring is tightly associated with the apoprotein. The protein-bound isoalloxazine ring does not exhibit large fluctuations on the nanosecond time scale, although small amplitude fluctuations cannot be excluded. This information was obtained by a combination of field-dependent 71 and NOE measurements. 7i values are in agreement with these results. On the basis of the dipolar part of the overall 71 values, the distance between the carbon investigated and the nearest proton was calculated and found to be in fair agreement with the crystallographic results of the related flavodoxin from Clostridium MP.
In addition, it is shown that, based on the chemical shift anisotropy as a relaxation source, information on the internal mobility is difficult to obtain. The main reason for this is the low precision in the determination of the chemical shift anisotropy tensor.
Internal motions in proteins have recently attracted considerable interest since the realisation that such motions play an important role in cooperative and allosteric effects. Moreover, the internal mobility in enzymes probably contributes to the catalytic effects of biomolecules [1]. Theoretical calculations by Karplus and McCammon [2,3] indicate that such motions can occur on the nanosecond —picosecond time scale despite the close-packed structure of native proteins. Such an internal mobility in the nanosecond time range has recently been demonstrated in the flavoprotein lipoamide dehydrogenase [4].
Flavoproteins occur widely in nature and catalyze a variety of biological reactions. The physical and chemical properties of the flavin prosthetic group are perturbed upon binding to a particular apoflavoprotein. These observations and model studies led to the postulation [5] that the molecular basis for the specificity of a particular flavoprotein could be a specific interaction between the apoprotein and the prosthetic group; i.e. hydrogen bond formation between some amino acid residues and certain atoms of the prosthetic group. That such hydrogen bonds do influence the orbital structure of the flavin molecule has recently been demonstrated by Eweg et al. [6], both by photoelectron spectroscopy and from theoretical calculations. Furthermore, it cannot be excluded that the prosthetic group in flavoproteins possesses an internal mobility which might be related to the particular function of a flavoprotein.
Ahhri'viuiions. FMN, riboflavin 5'-phosphate: Ac4rF. tetracetyl-ribotlavin: NMR. nuclear magnetic resonance: NOE, nuclear Overhauser effect: CSA. chemical shift anisotropy.
Flavodoxins are proteins of small relative molecular mass (A/r 15000 — 23000) and function as electron carriers in biological reactions. The prosthetic group of flavodoxins is riboflavin 5'-monophosphate (FMN). The one-electron redox potential of FMN is altered by complex formation with apo-flavodoxins [7]. Hydrogen bound formation between particular atoms of FMN and the apoprotein of Megasphaera elsdenii and Azotohaeter vinelandii flavodoxins has recently been demonstrated by 13C nuclear magnetic resonance (NMR) [8]. The question now arises whether the prosthetic group in flavodoxin is completely immobilized or still possesses some motional freedom. This uncertainty is related to the fact that the detailed mechanism of electron transfer is still poorly understood. It is possible that the degree of mobility of the prosthetic group in flavodoxins plays a role in the kinetics of electron transfer reactions.
Moonen and Müller [9], using the 31P NMR technique have recently demonstrated that the phosphate group of the protein-bound prosthetic group in M. elsdenii flavodoxin is tightly bound without any detectable internal mobility on the nanosecond time scale. Owing to the fact that the fluorescence emission of the prosthetic group in M. elsdenii flavodoxin is completely quenched, the mobility of the isoalloxazine moiety cannot be probed by nanosecond laser fluorescence techniques, i.e. time dependence of anisotropy. '3C relaxation measurements can also provide the desired information, however, and have been used to study the internal mobility of parts of proteins [10-12]. In these studies only hydrogen-carrying carbon atoms have been investigated. This has the advantage that one deals with a simple and well-defined
22
relaxation mechanism, i.e. the 13C-'H dipole-dipole relaxation mechanism.
With flavins, the situation is more complex because only C6 and C9 of the isoalloxazine ring carry protons and these positions are very difficult to enrich with 13C by chemical synthesis. We have therefore studied the flavins which are enriched at positions C2, C4, C4a and ClOa. Since the atoms in question are devoid of protons, analysis of our data had to include consideration of other possible relaxation mechanisms [13] as well.
The results show that it is possible, despite the complexity of our system, to obtain information on the mobility of the protein-bound prosthetic group by the combination of the following measurements: (a) spin-lattice relaxation at different magnetic field strengths, (b) nuclear Overhauser effects and (c) solid-state NMR to determine the principal elements of the chemical shift tensor.
MATERIALS AND METHODS
Flavodoxin from Megasphaera elsdenii (Lcl) was isolated and purified according to published procedures [14]. Apo-fiavodoxin was prepared as described by Wassink and May-hew [15]. The holoprotein was reconstituted at pH 7 — 8 by addition of an excess of flavin to the apoprotein followed by exhaustive dialysis of the solution.
Tetraacetylriboflavin (AtirF) was prepared from riboflavin as published earlier [16]. [2-13C]FMN, [4a-uC]FMN and [4,10a-l3C2]FMN were synthesized as described by van Schagen and Müller [8]. [4a-uC, 5-15N]riboflavin was synthesized from 7V-D-ribityl-3,4-dimethylaniline [17] by coupling with the diazonium salt of aniline using sodium [15N]-nitrite in the diazotation step. The resulting /V-ribityl-2-[15N]phenylazo-4,5-dimethylbenzene was condensed with [5-13C]barbituric acid as described previously [8].
13C NMR measurements were performed at 25.2 MHz on a Varian XL-100-15 NMR instrument equipped with a 16K620L computer, at 62.8 MHz on a Bruker WM 250 NMR spectrometer equipped with an Aspect 2000 computer and at 90.5 MHz on a Bruker HX 360 NMR apparatus equipped with a BNC 12 computer. All measurements were performed with quadrature detection and quadrature phase cycling, except for the measurements at 25.2 MHz, which were performed with single detection.
Proteins were dissolved in a mixture of 2H20/H20 (1 :9. by vol.) and buffered with sodium phosphate (100 mM. pH 7 — 8). The final concentration of flavodoxin in the NMR tube was 2 — 3 mM. AojrF was dissolved in C2HCli to a final concentration of 10 mM. FMN was dissolved in a mixture of 2H20/H2Û (1 :9, by vol.) or as otherwise stated. The solution was buffered with 100 mM sodium phosphate (pH 6.5). The final concentration of FMN was about 10 mM. In all experiments the temperature was kept at 26 + 2 =C. All samples were degassed.
At 25.2 MHz 12-mm and at 62.8 MHz and 90.5 MHz 10-mm Wilmad precision tubes were used, respectively.
Spin-lattice (7",) relaxation measurements were performed using the inversion-recovery method [18] as described by Levy and Peat [19]. At least 10 spectra were recorded for the determination of one 7"] value. The power of the broad-band decoupling was 2 W, except at 90.5 MHz: at this frequency, severe heating occurred at 2 W decoupling power and 1.2 W power was therefore applied. The experimental values were fitted to a single exponential function by computer analysis [20].
Nuclear Overhauser effect (NOE) values for the free flavins were obtained by integration of the resonance lines and comparing the values with and without proton decoupling. The waiting time in these experiments was at least four times T\ and the observation pulse was 90° as described by Lyerla and Levy [21]. NOE measurements for flavodoxin required long time averaging. For this reason, these measurements were performed with the help of a microprogram, in which alternate scans were obtained under continuous proton decoupling and under proton decoupling applied only during the detection period. Here again, waiting times of at least four times T\ were used. Subsequent subtraction of the free induction decays yielded the NOE values.
Spin-spin (T2) relaxation values were estimated from the line width of the experimental spectra.
Powder spectra were obtained on a Bruker CXP 300 NMR spectrometer. A special solid-state probe with a saddle coil was used. No proton decoupling was applied. A 90° pulse (5 us) was applied followed by a receiver dead time of 0.5 us and direct detection; 512 points were accumulated per free induction decay. Repetition times for the experiment are given in Results and Discussion. About 200 mg powdered riboflavin was used for the powder spectra and 20 — 200 scans were accumulated depending on the repetition time. Benzene was used as a reference. The benzene sample was also used to homogenize the magnetic field.
RESULTS AND DISCUSSION
The aim of this study is to obtain information on the internal mobility of the prosthetic group in Megasphaera elsdenii flavodoxin using 13C nuclear magnetic resonance relaxation. The description of the motional freedom is expressed by the spectral density function in the relaxation equations [22]. A precise knowledge of the spectral density function is necessary for the evaluation of the internal mobility. Different models have been proposed for the internal mobility [10,11,23]. The 'wobbling in cone' model, first proposed by Kinosita et al. [24], presents a reasonable description of the complex motions possible in proteins although one may argue that the internal motional freedom cannot be fully described by these restricted stochastic intramolecular motions. A possible internal motion of protein-bound flavin is illustrated in Fig. 1. The half-angle öm„ and the rotational coefficient TW
determine the internal motion, which is superimposed on the rotational motion of the whole protein. The choice of the
obbling
a: R=CH2-(CHOH)3-CH;OPOjH2
b: R=CH2-ICHOCOCH3)3-CH,OCOCH3
Fig. 1. The structure of (a) riboflavin 5'-phosphate and !bl tetraaeetyl-riboflavin. A model for the internal mobility of FMN bound to M. elsdenii flavodoxin is presented in dotted lines. The formalism of the "wobbling in cone' model is used [24]. The internal mobility is limited by a cone with the half-angle 0max. the frequency of the motion being expressed by the rotational coefficient rw. No full rotation around the wobbling axis is allowed. For further explanation see text
- 23
C2
ÏÉ fc JL
_ j ^
X tt JL
JU _A_
Jk__
2E " V £
" V
F £ "V
"TT-
" V
~ir~ i r
"T- f ?=
150 6 (ppm)
Fig. 2. Partially relaxed '3C AWÄ spectra of a mixture of equal concentrations of Act[2-"C/rF, Ac44a-xiC]rF and Ael[4,l0a-"Ci]rF in C2HCI3
measured at 90.5 MHz obtained by a 180J-T-90Ü pulse sequence and Fourier transformation
wobbling axis is arbitrary, but does not influence our final results within the limited accuracy of the model. In the case of an internal motion similar to that described in Fig. 1, Richarz et al. [10], following the 'wobbling in cone' model, arrived at this spectral density function :
J(W): 1 + (CUTC)2 • + ( 1 - 5 )
1 + (cury2 (1)
where S = '/* c°s2öma>, (1 + cos0m„)2 and (T^)_1 = TC-1
+ t~ ' (0.29 sin26m„)" '. TC is the rotational correlation coefficient describing the isotropic reorientational motion of the whole protein. If TW->GG orö„,a>->0, Eqn (1) becomes the normal spectral density function for an isotropic, rotational motion [22].
The 13C atoms (2, 4, 4a, 10a) studied in this paper are all non-protonated (cf. Fig. 1). From a theoretical point of view [25] the following relaxation mechanisms have to be considered for these carbon atoms: (a) the 13C-'H dipole-dipole interaction, (b) the 13C-14N dipole-dipole interaction and (c) the chemical shift anisotropy (CSA). The reader is referred to relevant textbooks [22,26] for the general equations of the relaxation by dipole-dipole interaction and chemical shift anisotropy. These two relaxation mechanisms have a different dependence on the spectral density function and the magnetic field strength. In order to obtain information on the spectral density function, it is therefore necessary to separate the relative contributions of the different relaxation mechanisms. One possibility is to evaluate the anisotropy of the chemical shifts of protein-bound FMN by analyzing the anisotropy in free FMN. 13C NMR measurements of FMN, free in aqueous solution and bound to M. elsdenii flavodoxin, showed that, despite certain differences due to specific hydrogen
bonding interactions, the chemical shifts do not differ much [8]. This strongly suggests that the anisotropy of the chemical shifts in free FMN is indeed roughly the same as in protein-bound FMN. A disadvantage in using free FMN is the formation of aggregates in aqueous solutions [9]. We have therefore also studied tetraacetylriboflavin (Ac4rF) in chloroform, which does not form aggregates in this solvent [27], The change of solvent (water to chloroform) causes certain differences in polarization of the aromatic ring as evidenced by 13C NMR [8], but the chemical shifts do not differ much on an absolute scale. Thus, the use of free FMN in aqueous solutions and Ac4rF in chloroform is allowed (with some caution) for an evaluation of the anisotropy of the chemical shifts of protein-bound FMN.
Free-Flavins in Solution
Theoretically it is possible to determine the chemical shift anisotropy using field-dependent relaxation measurements [28 — 30]. Representative results of relaxation measurements are presented in Fig. 2. The 13C relaxation data of Ac4rF in chloroform are collected in Table 1. The overall T, values of Ac4rF at 25.2 MHz have already been estimated by Ghisla et al. [31] and are in fair agreement with our data. In the motional narrowing limit (COTC < 1) the relaxation by 13C-'H dipole-dipole interaction (TP0-") can be calculated from the overall 7\ and NOE values [32]. From Table 1 it can be concluded that the Tf1^" values are independent of the magnetic field strength. This can be taken as proof that we are dealing with the motional narrowing limit in both magnetic field strengths [32]. In the motional narrowing limit 7fSA is the only relaxation parameter influenced by the magnetic field strength [28]. Thus, from a simple comparison of
24
Table 1. The dependence on the magnetic field .strength of the overall spin-lattice relaxation time* (T,), nuclear Overhauser effects (NOE) and the calculated relaxation due to dipole-dipole interaction (T^D^H) of some l3C atoms of tetraacetylriboflavin in C2/7C/3
The precision of the T\ values is + 5 %, that of the NOE values is ± 0.05, and that of T"D~H is dependent on the NOE value (see text for further explanations)
C atom Frequency 7"i NOE 7?r,
C2 C4 C4a ClOa
C2 C4 C4a ClOa
MHz
90.5 90.5 90.5 90.5
25.2 25.2 25.2 25.2
5.1 5.8 5.7 4.7
11.3 12.1 35.9 16.4
1.45 1.48
< 1.10 1.10
1.89 2.06
< 1.10 1.69
22.4 23.9
> 5 0 46.3
25.3 22.6
^ 5 0 47.2
Table 2. The relaxation clue to chemical shift anisotropy (T\SA) of some 13C atoms in tetraacetylriboflavin (Ac*rF) in C2HC/s and riboflavin 5'-monophosphate (FMN) in aqueous solution at 90.5 MHz and the calculated values for the anisotropy in the chemical shift ( Aa j The precision of the 'T\%* values for Ac4rF is ±10% and that for FMN about + 20",,. For the accuracy of the Ao values, see text
Compound
A c r F in C 2 HCb
FMN i n 2 H z O
"C atom
C2 C4 C4a ClOa
C2 C4 C4a ClOa
Tf"
8.5 10.2 6.3 6.0
6.6 4.8 3.4 2.7
Ao
205 185 239 245
161 188 224 250
the 7"i values at different magnetic field strengths the 7"iCSA
can be calculated [32] (Table 2). It is evident from Tables 1 and 2 that both the CSA and the 1 3C-'H dipole-dipole relaxation mechanisms are important, especially at high magnetic field strength. In contrast, other relaxation mechanisms such as the 1 3C-14N dipole-dipole interactions contribute little to the overall relaxation rates.
A similar relaxation study was done on FMN in H 2 0 . Treatment of these data obtained at two field strengths (not shown) in analogy to those of Ac4rF lead to the same conclusions as mentioned above for Ac4rF, i.e. the CSA relaxation plays an important role in the overall relaxation of the carbon atoms studied. The overall T\ values of F M N in 2 HiO are summarized in Table 3. In D 2 0 the proton attached to N3 of F M N (cf. Fig. 1) exchanges easily with a deuteron of the solvent. The NOE values are therefore very small in this solvent. It should be noted that the accuracy of the results for F M N in 2 H 2 0 or H 2 0 is lower than that for Ac4rF in CHCI3 because FMN contains impurities consisting of monophosphate isomers and riboflavin and also forms aggregates [9], The 7fSA values calculated for FMN are presented in Table 2.
Now the T\%K values are known, the CSA parameters Aa can be evaluated. The Aa term is composed of the principal elements a,, (1 = v, r, r) of the chemical shift tensor:
l a = a,._-'/2 (a ,v + <j„). (2)
Table 3. The dependence on the magnetic field strength of the overall spin-lattice relaxation times ('T,;, of some ,3C atoms of FMN in 2tÎ20 The precision of the T{ values is + 10%
13C atom 7", at
90.5 MHz 25.2 MHz
C2 C4 C4a ClOa
3.5 3.3 2.2 1.8
6.9 7.9 5.2 4.3
The general equation for 7 |S A [22] also contains the term rj, which expresses the deviation of the chemical shifts tensor from axial symmetry [33].
•1 • ffiv
(3)
(4)
where aav is the average (liquid-state) chemical shift.
<r„v = Va ( ' „ + %• + <*»)•
The CSA parameter Act of the investigated carbon atoms of the isoalloxazine moiety of Ac4rF and F M N can now be calculated from the 7^SA values [28 — 30]. For an asymmetric flat aromatic ring system as present in Havin we can expect an anisotropic rotational diffusion tensor. The only information on the reorientational motion can be obtained from the 7?°"" values of C4 and C2, arising mainly from the interaction of these carbon atoms with the proton at N3 of flavin. In fact, however, this information represents only one term of the rotational diffusion tensor. In order to obtain an estimation of the Aa values, it is necessary to assume that the reorientational motion is isotropic, i.e. J(co) = TC in the motional narrowing limit. As we know the distance between the carbon atoms C2, C4 and the N3 proton (0.2 nm) from crystallographic studies [34], rc can easily be calculated. A rc
of 74 ps is calculated for Ac4rF in chloroform and a tc of 165 ps for F M N in H 2 0 at 26 =C. The difference between these values corresponds well with the difference in viscosity between the two solvents at 26 ; C. To estimate the Aa values using Eqn (2) the additional assumption has to be made that the chemical shift tensor is axially symmetric, i.e. axx = ayy
and therefore 1; = 0 (Eqn 3). The calculated Aa values for Ac4rF and F M N are given in Table 2. The order of magnitude of the Aa values is about the same as that reported for indole [30] and for benzene [35]. In addition the Aa values for FMN and Ac4rF are quite similar (Table 2). This result was expected since the electronic structure of the isoalloxazine moiety of F M N and Ac4rF does not change drastically on going from water to chloroform. The largest difference observed is between C2 in Ac4rF and in FMN. This difference probably reflects the greater polarization of the carbonyl group at position 2 in aqueous solution as compared to that in CHCI3. This suggestion is supported by previously published work [8].
Solid-Stali' NMR of Flavin
Solid-state N M R was applied to powdered riboflavin to test the validity of the two assumptions made above. Powder spectra enable determination of the principal elements of the chemical shift tensor [33,36] and therefore provide a very
25
Fig. 3. 13C solid-state NMR spectrum of powdered [4a-iiC, 5-l5NJribo-flavin at 20°C. The repetition rate was 1 h. Total number of scans is 20. For the numbering of the peaks see text
straight-forward way to determine CSA parameters An and v] using Eqns (2) and (3). This method can however only be applied for the chemical tensor of C4a of flavin. The reason for this is that C4 and ClOa in flavin cannot be enriched independently and a doubly enriched flavin would yield a very complex powder spectrum. In addition, neighbouring quadrupolar nitrogen atoms (cf. Fig. 1) would make such powder spectra even more complex [37]. Synthesis of a flavin derivative enriched with 13C at position 2 and 15N at positions 1 and 3 would have been too expensive for the limited use of the compound. On the other hand, it is relatively easy in riboflavin to enrich C4aby 13C and N5 by 15N. The nearest hydrogen atom to the quaternary C4a atom in flavin is three bonds away. Consequently, cross-polarization from proton magnetization with or without 'magic-angle' spinning would probably be useless. For this reason, the experimental approach for this problem was the use of a single 90 pulse followed by a very short receiver dead time and acquisition (cf. Materials and Methods). In Fig. 3 a typical powder spectrum of [4a-13C, 5-15N]ribofiavin is shown. Although the spectrum is rather complex a number of peaks and shoulders can be distinguished. We can expect very large T\ values (perhaps even exceeding an hour) for non-protonated carbon atoms in powders. As riboflavin also contains protonated carbon atoms (characterized by much smaller T\ values) we can expect intensity from some protonated carbon atoms, besides the intensity from the enriched C4a atom (Fig. 3). The differences in 7\ values mentioned above should however provide a method for the assignments. This was achieved experimentally by varying the repetition time between 1 h and 10 s. In this way it is possible to estimate roughly the T\ values: the peaks at 20 ppm (peak 6), 60 ppm (peak 5) and 120 ppm (peak 2) have relatively short 7*i values. In liquid-state spectra of flavins [38] the C7 and C8 carbon atoms, the C atoms of the ribityl side chain and the C6 and C9 atoms are found respectively at these frequencies. The peaks 1, 3 and 4 in Fig. 3 exhibit a very long Tt and this suggests that these peaks represent the principal values of the chemical shift tensor of the C4a atom. These tensor elements, i.e., CT„. <T,.,. and a.-;, are at 75 ppm, 104 ppm and 217 ppm. Calcula
tion of the average chemical shift <jav (Eqn 4) gives 132 ppm which is in good agreement with the value of 136 ppm observed in the liquid state [8,38]. Using Eqns (2) and (3), Aa and i) are calculated as 127 ppm and 0.34, respectively. This result differs from the values in Table 2 by a factor of almost 2. Although the assignments of the tensor elements should be regarded with caution, the results clearly show that the Aa values obtained in the previous section from liquid-state relaxation measurements are overestimated. The reason for this is that the above-mentioned assumptions are too rigorous. Thus we have only obtained a rough estimation of the CSA parameters for the C4a atom. Consequently the CSA values of Table 2 cannot be applied to a relaxation study of protein-bound FMN. Nevertheless, we will show that it is possible to obtain unambiguous information on the internal mobility of protein-bound FMN by other NMR data (cf. below).
Protein-Bound FMN
The rotational correlation time of the protein must be known to analyze the relaxation data on flavodoxin from M. elsdenii. The physical and chemical properties of M. els-denii flavodoxin are very similar to those of the related flavodoxin from Clostridium MP.. Crystallographic studies on the latter protein revealed that it possesses an almost ideal globular structure [39]. These facts suggest that M. elsdenii flavodoxin also possesses a globular structure. This allows us, knowing that the translational diffusion coefficient D20 w
= 11.4x10"7cm2 s ' [14], to calculate 5.9 ns format 26'C. On the other hand, a rc of 5 ns is calculated using the Stokes-Einstein relation and considering a hydration shell of the protein. Since the NMR measurements require a relatively high concentration of protein, we determined the macroscopic viscosity for a 3 mM solution of flavodoxin and found that the viscosity increases less than 10% at 26 C as compared to that of a dilute solution. It was shown [9] by analytical ultracentrifugation experiments that even at these concentrations no aggregation occurs. The two independently determined TC values are thus in good agreement. In the following we will use the TC value of 5 ns, since, as will be shown below, the conclusions drawn from the relaxation data do not depend critically on the tc value used in the calculations. A representative '3C T\ measurement for selectively 13C-enriched FMN bound to the apoflavodoxin from M. elsdenii is shown in Fig. 4. The 13C resonances appear as doublets due to double labelling of FMN, i.e. C4 and ClOa [8]. The spin-lattice relaxation times, measured at 62.8 MHz and 90.5 MHz, are collected in Table 4.
Theoretically [25], and also as deduced from model studies, we can expect the CSA and the dipole-dipole (DD) interactions to be possible relaxation sources. Let us first consider the theoretical dependence of if0 and 7fSA on the isotropic TC value. The theoretical curves are shown in Fig. 5. Fig. 5B leads to the deduction that 7fSA values are smaller at high field strength than at low field strength. In addition, 7fSA
gradually becomes less dependent on the field strength with increasing TC and for onc §> 1, 7rSA is independent of the field strength. The opposite is true for T°D (Fig. 5A). In Fig.5A only the 13C-'H dipole-dipole relaxation is shown. The curves for 13C-14N dipole-dipole relaxation are similar to those of 13C-'H dipole-dipole relaxation [41] however. These theoretical data, in combination with overall field-dependent Tj values of protein-bound FMN (Table 4), allow us to draw some important conclusions: (a) the observed increase in the overall T\ values with increasing field strength
- 26 -
•JV^k^VI
:^/v^AV V Wr
1 0.01
I 0.1
1 0.2
1 0.«
1 0.7
1 1.0
T Isl
1 1.4
1 1.9
1 2.5
1 3.5
1 5.0
Fig. 4. 13C Ti and NOE measurement at 62.8 MHz. Partially relaxed , 3 C-NMR spectra of [4,10a-13C2] FMN bound to M.elsdenii flavodoxin measured at 62.8 MHz obtained by a 180 -r-90° pulse sequence and Fourier transformation
Table 4. The overall spin-lattice relaxation times (~[\) of certain atoms of selectively liC-enriched FMN bound to the apoflavodoxin from M. elsdenii measured at two different frequencies
13C atom
C2 C4 C4a ClOa
r, at
62.8 MHz
s
1.5 + 0.1 1.5 + 0.05 1.0 + 0.1 1.2 ± 0.05
90.5 MHz
1.9 + 0.15 2.1 + 0.1 1.4 + 0.15 1.8 ± 0.1
must be due to the dipole-dipole relaxation mechanism; (b) the observed increase in 7i values at high field strength must be due to a significant contribution of slow motions of protein-bound FMN (i.e. iOTc> 1) to the spectral density function. Thus, a very mobile protein-bound FMN can already be excluded. In the following we will analyze the data of Table 4 in detail. Calculations show that for a TC of 5 ns and assuming no internal mobility 7?D increases by a factor of 1.88 on going from 62.8 MHz to 90.5 MHz. As can be checked easily the value of 1.88 is independent of the relative contributions of 13C-'H and 13C-14N to the T°D value. If an internal mobility on the nanosecond or subnanosecond timescale exists, the factor of 1.88 will decrease (cf. Fig. 5A). The 7fSA value will always decrease on going from 62.8 MHz to 90.5 MHz (cf. Fig. 5B). Consequently, the magnitude of the increase in overall 7"i values for all four carbon atoms at high field strength allows the determination of the lower limit of the contribution of the dipole-dipole interactions to the overall Tt relaxation. As an example, an increase in the overall Tt from 1.5 s to 2.1 s is observed for the C4 atom (Table 4). Only dipolar interactions can explain this increase.
Knowing that the ratio of T?D (90.5 MHz)/7?D (62.8 MHz) is smaller than or equal to 1.88, we simply calculate that T?D makes up at least 60% of the overall T, relaxation at 62.8 MHz. In analogy a lower limit of the relative contribution of the dipolar interactions is determined as 45% for C2, 62% for C4a and 71% ClOa. Both the 13C-lH and the 13C-14N dipole pairs contribute to T?D. The situation is simple for C2 and C4, as the proton attached to N3 is very close (0.2 nm) and accordingly the 13C-14N dipolar pairs can be safely neglected. A more complex situation exists for C4a and ClOa as the closest hydrogen atoms are probably at van der Waals distances. Owing to the small values of 7?D however, the contribution of the 13C-14N dipole interactions to T?D is probably also minor for these carbon atoms. Summarizing, the overall Tt values of C2 and C4 have a 7?°"" contribution of at least 45% and 60%, whereas for C4a and ClOa the contribution of T?°~H is not known exactly, but is probably at least 50%. This means that 13C-'H nuclear Over-hauser effects should yield valuable information. This is illustrated for the C4 atom where 7?°"" makes up at least 60% of the overall 7, value at 62.8 MHz. As the 13C-'H NOE only deals with the 13C-'H dipole interactions, the NOE will be (depending on the mobility) between 1 and at least 2.2 (60% of the maximal increase). Thus an NOE measurement should give the desired information on the internal mobility. Fig. 5C demonstrates that only a very small NOE will be detected at 62.8 MHz if no internal motion is present. A considerable contribution from fast internal motions (T„ < 1 ns) will enhance the NOE substantially however.
We have carried out NOE measurements at 62.8 MHz. No NOE could be detected for the carbon atoms investigated. For at ze of 5 ns we calculate an NOE of 1.19 at 62.8 MHz if the 13C-'H dipole-dipole interaction is the only relaxation source. Since the minimal contribution of the 13C-'H dipole-dipole interaction is 60% for C4, the NOE at 62.8 MHz for C4 is between 1.11 and 1.19 for a rc of 5 ns and no internal motion. Within the accuracy of the measure-
27
Chemical shift anisotropy
3 |
v/-
Fig. 5. The theoretical dependence of (A) the dipole-dipole, {B) the chemical shift anisotropy relaxation* and (CJ the nuclear Overhauser effect on the isotropic rotational correlation time frcJ is shown for the case that Eqn (II holds for the spin density function (see text). The dependence is calculated for two different field strengths as indicated
merits the experimentally observed NOE is smaller than 1.15. As no NOE is detected for the other investigated carbon atoms either, we can state that the bound isoalloxazine moiety contains no considerable internal motion in the nanosecond or subnanosecond timescale. We can also see from Fig.5C that this conclusion does not depend critically on the accurate knowledge of the overall rotational correlation time of the flavodoxin, as the NOE in the presence of a substantial motional freedom in the nanosecond or subnanosecond region would be hardly dependent on the exact overall rotational correlation time. We can discuss in more detail the internal motions which are possible in the light of our results (cf. Eqn 1). Our results cannot exclude an internal motion characterized by a r» > 3 ns. On the other hand, an internal motion characterized by a T» < 1 ns is only possible when the half-angle 0m„ is smaller than 20 . In this case the contribution of the internal motion to the spectral density function is too small to allow its detection by the current limitation of the instruments.
The conclusion that no internal motion contributes significantly to the overall relaxation time allows us to calculate
Tf°~H from the field-dependent measurements and therefore also the distance between the 13C and the nearest hydrogen atom(s). The distance to the nearest hydrogen atom is known for C2 and C4, i.e. the hydrogen atom attached to N3. The distance is 0.2 nm for both carbon atoms as deduced from crystallographic data of flavins [34]. Distances of 0.19 nm between C2 and its nearest hydrogen atom and 0.18 nm between C4 and its nearest hydrogen atom are calculated from our data. As hydrogen atoms in van der Waals' contact might contribute to a small extent to the jfD-H relaxation the calculation might yield an underestimation of the distances. Either way, the calculated distances are in good agreement with the crystallographic data [34].
The conclusion that the isoalloxazine ring is immobilized by the apoprotein can be checked on the basis of the CSA parameters of the C4a atom, i.e. Aa (127ppm) and t\ (0.34). Using these data, the overall 7\ values of the C4a atom of protein-bound FMN should be in agreement with a spectral density function in which only the overall reorientational motion of the protein is accounted for. A 7fSA of 2.81 s is then calculated for C4a at 62.8 MHz. As the overall T\ is 1.0 s (Table 4), we calculate a T?D of 1.55 s at 62.8 MHz. Using the theoretical dependence of 7?D and 7fSA on the magnetic field strength, TfD and 7fSA are calculated as 2.91 s and 2.51 s at 90.5 MHz, respectively. From these values an overall Tt of 1.35 s is calculated. This value agrees well with that determined experimentally (1.4 s, Table 4). Consequently, this supports the conclusion that the protein-bound isoalloxazine ring does not have a significant motional freedom which is characterized by a t„ < 1 ns.
Also T2 values were determined from linewidth (zl Vi/2) measurements (7i = l/27r^v1;2). The linewidth was 2 + 0.7 Hz for all four carbon atoms, independent of the field strength. These values are in accordance with our conclusions given above, but the inherent inaccuracy does not permit additional conclusions.
From this study and results published previously on 3 JP NMR on M. elsdenii flavodoxin [9], it must be concluded that the prosthetic group in flavodoxin from M. elsdenii is tightly associated with the apoprotein. This is an interesting observation with regard to nanosecond fluorescence data on the pyruvate dehydrogenase complex [42] for which it was found that the protein-bound flavin possesses a certain internal freedom on the nanosecond timescale. It is possible that these observations are linked to the specific biological functions of the proteins under consideration.
We are indebted to Mr M. M. Bouwmans for the preparation of the figures, Mrs J. C. Toppenberg-Fang for typing the manuscript, Mr W. A. M. van den Berg for the isolation of the flavodoxin. Dr C. A. H. Rasmussen for linguistic advice and Dr J. Lohman (Bruker. Wormer, The Netherlands) for help with the measurements at 62.8 MHz. This study has been carried out under auspices of the Netherlands Foundation for Chemical Research (S.O.N.) with financial aid from the Netherlands Organization for the Advancement of Pure Research (Z.W.O.).
REFERENCES
1. Gready, J. E. (1979) in Catalysis in Chemistry and Biochemistry: Theory and Experiment (Pullman. B.. ed.) pp. 11 - 2 3 , D. Reidel Publishing Co.. Dordrecht.
2. Karplus. M. & McCammon. J. A. (1979) Nature (Land.) 277, 578. 3. Karplus. M. & McCammon. J. A. (1981) CRC Cril. Rev. Biochem.
9. 2 93-349 . 4. Visser. A. J. W. G.. Grande. H. J. & Veeger. C. (1980) Biophys.
Chem. 12. 3 5 - 4 9 .
28
5. Müller, F. (1972) Z. Naturforsch. 27b, 1023-1026. 6. Eweg, J. K., Müller, F., van Dam, H., Terpstra, A. & Oskam, A.
(1982) J. Phys. Chem. 86, 1246-1251. 7. Mayhew, S. G. & Ludwig, M. L. (1975) in The Enzymes (Boyer,
P. D.,ed.) 3rded, vol. 12, pp. 57—118, Academic Press, New York. 8. Van Schagen, C. G. & Muller, F. (1981) Eur. 7. Biochem. 120,
33-39. 9. Moonen, C. T. W. & Müller, F. (1982) Biochemistry, 21, 4 08 -414 .
10. Richarz, R„ Nagayama, K. & Wüthrich, K. (1980) Biochemistry, 19, 5189-5196.
11. Jardetzky, O., Ribeiro, A. A. & King, R. (1980) Biochem. Biophys. Res. Commun. 92, 8 83 -888 .
12. James, T. L. (1980)7. Magn. Reson. 39, 1 41 -153 . 13. Norton, R. S., Clouse, A. O., Addleman, R. & Allerhand, A. (1977)
J.Am. Chem.Soc. 9 9 , 7 9 - 8 3 . 14. Mayhew, S. G. & Massey, V. (1969) 7. Biol. Chem. 244, 7 94-802 . 15. Wassink, J. H. & Mayhew, S. G. (1975) Anal. Biochem. 68, 6 09 -616 . 16. Müller, F. (1971) Methods Enzymol. I8B, 4 53 -458 . 17. Lambooy, J. P. (1971) Methods Enzymol. I8B, 4 37 -447 . 18. Void, R. L„ Waugh, J. S., Klein, M. P. & Phelps, D. E. (1968) 7.
Chem. Phys. 48, 3881-3832. 19. Levy, G. C. & Peat, I. R. (1975) 7. Magn. Reson. 18, 5 0 0 - 521 . 20. Sass, M. & Ziessow, D. (1977) 7. Magn. Reson. 25, 2 63 -276 . 21. Lyerla, J. R. & Levy, G. C. (1974) in Carhon-13 NMR Spectro
scopy (Levy, G. C ed.) pp. 79— 148, John Wiley and Sons, New York.
22. Abragam, A. (1961) The Principles of Nuclear Magnetism, pp. 264 — 323, Clarendon Press, Oxford.
23. Woessner, D. E. (1962) J. Chem. Phys. 36. 1 - 4 . 24. Kinosita, K., Jr. Kawato, S. & Ihegami. A. (1977) Biophys. J. 20,
289-305 . 25. Allerhand, A. (1979) Methods Enzymol. 61. 4 58 -549 .
26. Farrar, T. C. & Becker, E. D. (1971) Pulse and Fourier Transform NMR, pp. 46 — 65, Academic Press, New York.
27. Kyogoku, Y. & Yu, B. S. (1968) Bull. Chem. Soc. Japan, 41, 1742. 28. Wehrli, F. W. (1978)7. Magn. Reson. 32, 4 51 -454 . 29. Schweitzer, D. & Spiess, H. W. (1974)7. Magn. Reson. 16. 2 4 3 - 251 . 30. Levy, G. C. & Edlund, U. (1975)7. Am. Chem.Soc. 97 ,5031-5032. 31. Ghisla, S„ Hastings, J. W„ Favaudon, W. & Lhoste, J. M. (1978)
Proc. Natl Acad. Sei. USA, 75, 5860-5863. 32. Wehrli, F. W. (1976) in Interpretation of Carbon-13 NMR Spectra
(Wehrli,F.W. & Wirthlin.T., eds) pp. 129 -151 , Heyden, London. 33. Haeberlen, U. (1981) Phil. Trans. R. Soc. Lond. A299, 4 97 -504 . 34. Kierkegaard, P., Norrestam, R., Werner, P.-E., Csöregh, I., von
Glehn. M., Karlsson, R., Leijonmarck, M., Rönnquist, O., Stems-land, B., Tillberg, O. & Torbjörnsson, L. (1971) in Flavins and Flavoproteins (Kamin, H., ed.) pp. 1—22, University Park Press, Baltimore.
35. Pines. A., Gibby, M. G. & Waugh, J. S. (1972) Chem. Phys. Lett. 75 ,373-376 .
36. Kempf, J., Spiess, H. W., Haeberlen, U. & Zimmermann, H. (1974) Chem. Phys. 4 , 2 69 - 276 .
37. Van der Hart, D. L., Gutowsky, H. S. & Farrar, T. C. (1967) 7. Am. Chem. Soc. 89, 5056-5057.
38. Grande, H. J., Gast, R„ van Schagen, C. G , van Berkel, W. J. H. & Müller, F. (1977) Helv. Chim. Acta, 60, 3 67-379 .
39. Burnett, R. M„ Darling, G. D., Kendall, D. S., LeQuesne, M. E., Mayhew, S. G., Smith, W. W. & Ludwig, M. L. (1974) 7. Biol. Chem. 249, 4383-4392.
40. Visser, A. J. W. G., Scouten, W. H. & Lavalette, D. (1981) Eur. 7. Biochem. 121, 2 33 -235 .
41. Norton, R. S. & Allerhand, A. (1976) 7. Am. Chem. Soc. 98, 1 007 -1014.
42. Grande, H. J., Visser, A. J. W. G. & Veeger, C. (1980) Eur. J. Biochem. /0Ö, 361 —369.
C. T. W. Moonen and F. Müller. Laboratorium voor Biochemie der Landbouwhogeschool, De Dreijen 11, NL-6703-BC Wageningen, The Netherlands
- 29 -
Chapter 5
A REINVESTIGATION OF THE STRUCTURE OF OXIDIZED AND REDUCED FLAVIN.
13 15 A C and N nuclear magnetic resonance study
Chrit T.W. Moonen, Jacques Vervoort and Franz Müller
- 30 -
SUMMARY
Several chemically substituted flavins are investigated in the oxidized
13 15
and the reduced state by C and N NMR techniques. The dependence on the pola
rity of the solvent and on the concentration are studied. In combination with
already published results a semi-empirical theory is developed to interpret
the chemical shifts in terms of the solution structure of flavins. Where pos
sible the results are compared with crystallographic data. The two sets of data
appear to be in good agreement.
In contrast to common ideas the oxidized state is not fully coplanar, but
the N(10) atom is situated out of plane to a certain degree. Polarizing the fla
vin by hydrogen bonds in a high dielectric medium moves the N(10) atom into the
molecular plane and the flavin molecule becomes coplanar. In the coplanar molec
ule IT electrons are delocalized from the N(10) atom mainly to 0(2a) and 0(4a).
The NMR results show that the solution structure of reduced flavin is mainly
governed by sterical hindrance and hydrogen bonds. The findings are in contrast
to commonly accepted ideas that reduced flavin is strongly bent. In an apolar
solvent the reduced neutral isoalloxazine is only slightly bent. The formation
of hydrogen bonds in a protic solvent of a high dielectric constant decreases 2
the bend. The N(10) atom is now almost fully sp hybridized and the N(5) atom
has an endocyclic angle of 115°-117°, indicating its predominant sp character.
The results have several important implications for flavin catalysis. Among
these, it is shown that the altered redox potential of the semiquinone-fully
reduced redox couple of flavodoxin is probably not caused by the planarity of
the reduced protein-bound FMNH".
INTRODUCTION
The flavins are especially remarkable among the various known natural redox
coenzymes because of two outstanding features: i) they can function as one-
electron as well as two-electron redox carriers, ii) they act with considerable
- 31 -
efficiency in a wide variety of enzymatic reactions. As a consequence, the fla
vins are known as very versatile redox coenzymes. This versatility suggests that
nature has many possibilities to 'tune' the function of the flavin. This has led
to the proposal that specific interactions between flavin and apoflavoprotein
play a particular role in determining the pathway of flavin catalysis (Müller,
1972; Müller et al_., 1970; Hemmerich and Massey, 1982). Other factors such as
mobility of the flavin (Moonen and Müller, 1983), microenvironment of the flavin
binding site as well as the planarity of the reduced flavin molecule (Tauscher
et a]_., 1973; Simondsen and Tollin, 1980) may also be important.
Nuclear Magnetic Resonance is a powerful tool to test these hypotheses as,
in principle, the modulation of the structure and electron density can be moni-
13 15 tored using C and N NMR techniques (Van Schagen and Müller, 1981; Franken
et a K , 1983). Moreover, dynamic information can also be obtained by the NMR
method on various time scales (Moonen and Müller, 1983; Moonen and Müller, 1982a;
Moonen et aj_., 1982b). A detailed 1 5N (Franken et al_., 1983) and 1 3C (Van Schagen
and Müller, 1981) NMR study of M.elsdenii flavodoxin actually showed that some
specific interactions could be unambiguously assigned. No satisfying explanation
13 15 could be given however, for some remarkable chemical shifts in C and N spectra.
Among these, we mention the chemical shift of the N(10) atom of protein-bound FMN
in both the oxidized and the reduced state. The chemical shifts due to C(10a) and
C(4a) could not be explained satisfactorily either. The difficulties in the inter
pretation of the chemical shifts exist not only for protein-bound flavins but
13 15
also for free flavins. Although extensive C and N NMR studies have been car
ried out on free flavins (Grande et a K , 1977a; Van Schagen and Müller, 1980;
Yagi et al_., 1976; Kawano et al_., 1978), some chemical shift changes still remain
unexplained. Among these are the chemical shifts due to N(10), C(10a) and C(4a)
as well as C(7), C(5a), C(9a), both in the oxidized and reduced flavin. The dif
ficulties in the interpretation probably originate from the various possibilities
to modify the structure and the electron density of flavins by the medium used,
which might be related to the biological versatility of flavins. We report here
a reinvestigation of oxidized and reduced flavins by NMR techniques, using
- 32 -
various chemically modified flavins. The aim of this detailed study is to detect
the various possibilities of modulating the physical and chemical properties of
flavins and to pre
on flavoproteins.
13 15 flavins and to provide a basis for the interpretation of C and N NMR results
MATERIALS AND METHODS
13 15 C and N substituted flavins were prepared as described previously
(Van Schagen and Müller, 1981; Moonen and Müller, 1983; Müller et aj_., 1983).
Tetraacetylri boflavin was prepared from riboflavin by acetylation in a mixture
of CH3C00H/(CH3C0)20 in the presence of a small amount of HC104 (Müller, 1971).
Phosphorylation of riboflavin derivatives was done according to the procedure
of Scola-Nagelschneider and Hemmerich (1976). The synthesis of all other com
pounds has been described previously (Grande et a K , 1977b; Van Schagen and
Müller, 1980).
Wilmad 10 mm precision tubes were used. The samples contained about 3 mM
13 15 flavin, if C and N enriched compounds were used, unless otherwise stated.
13 15 Natural abundance C and N NMR measurements were performed with samples
containing 20-200 mM flavin. The sample volume was 1.6 ml. Aqueous samples con
tained 10% 2H20 for 1 3C NMR and 100% 2H20 for 1 5N NMR (to exchange NH protons
for deuterons). pH Measurements were done before and after the NMR measurements
and are reported without correction for isotope effects. Aqueous samples con
tained 100 mM potassium phosphate, pH 8.0. Reduction was conducted by the addi
tion of the desired amount of a dithionite solution to the anaerobic solutions.
Anaerobiosis was achieved by carefully flushing the solutions in the NMR tube
with argon for about 20 min. The NMR tube was sealed with a serum cap. Reduction
2
of oxidized flavin solutions in C H C K to the 1,5-dihydro state was effected di
rectly in the NMR tube by vigorous shaking of a two-phase solution consisting
of the flavin solution in C HCl3 and an aqueous solution of 0.5 M potassium
phosphate (pH 8.0, saturated with KCl) containing a 10-fold excess of sodium
dithionite with respect to the flavin.
- 33 -
Measurements were performed on a Bruker CXP 300 NMR spectrometer operating
at 30.4 MHz for 1 5N and 75.6 MHz for 1 3 C NMR measurements. Broadband decoupling
13 15
of 2 Watt was applied for C NMR. No decoupling was applied for N NMR. The
temperature of all samples was kept constant at 26 ± 2°C. All spectra were record
ed using 30° pulses and a repetition time of 0.8 - 1.3 s. Dioxan was used as an 2 13
internal reference in aqueous solution and TMS in C HCl, solutions for C NMR ^dioxan " 5TMS = 67-8^ PPm)- Chemical shifts are expressed relative to TMS. Neat 15 15
[ N]CH3N02 was used as an external reference in all solutions for N NMR using
a coaxial cylindrical capillary (Witanowski et al_., 1981). Chemical shifts are
reported as true shieldings (i.e. corrected for bulk volume susceptibilities)
relative to liquid NH3 (6~„ N0 - SN„ = 381.9) (Witanowski et al_., 1981;
Levy and Lichter, 1979).
RESULTS AND DISCUSSION
13 C NMR studies of substituted benzene derivatives have provided empirical
justification for the idea that changes in ir electron densities are primarily
responsible for the observed shielding variations (Lauterbur, 1962). Karplus
and Pople (1963) developed a theory which supports this notion. Lauterbur (1965)
showed that simple aromatic heterocycles also follow this rule. Pugmire and
Grant (1971) investigated more complex aromatic heterocycles and concluded that
the parameters derived from simple heterocycles can be satisfactorily applied.
Up till now several studies confirmed that the semi-empirical relationship be-
13 tween IT electron density and C chemical shift in aromatic systems are valid
(Levy and Lichter, 1979 and references therein).
15 The same rules can be applied for N chemical shifts, although with some
caution with regard to the marked differences between pyridine- and pyrrole
type nitrogens (Witanowski et al_., 1981, and references therein). The explana-
15 tion of the N chemical shift of a deprotonated nitrogen atom in heteroaromatic
systems is theoretically more difficult to explain. The investigated systems
always follow the same rules however, in that the deprotonated atom shifts con-
- 34 -
siderably downfield, although the electron density is increased (Pugmire and
Grant, 1968, 1971; Quirt et a1_., 1974). These studies showed in addition that
the atoms in a position to the deprotonation site show also a downfield shift,
for all other atoms the IT electron increase or decrease governs the chemical
shift.
Van Schagen and Müller (1980) showed that, assuming a "butterfly" conforma
tion of reduced flavin along the N(5) - N(10) axis, a good agreement exists be-
13 tween the IT electron density and the C chemical shift of a particular atom.
In the oxidized state, assuming a flat aromatic system, the agreement is less
satisfying and this fact could not be explained at that time.
In the interpretation of our results we consider, taking into account the
above-mentioned semi-empirical rules, the following parameters (for a discussion
see Levy and Lichter, 1979; Witanowski et al_., 1981): i) the degree of hybridi
zation, ii) the IT electron density, iii) the different behaviour of pyridine-
and pyrrole-type nitrogen atoms and iiii) the influence of a negative charge on
the deprotonation site and on the neighbouring atoms (only relevant for the inter
pretation of anionic reduced flavins).
9xlC!l?Cd_flavin
The oxidized isoalloxazine ring of flavin has commonly been assumed to be a
flat aromatic ring system. It is generally accepted that the carbonyl function
at C(2) plays a major role when the molecule is polarized in polar solvents
(structure b, Figure 1). In keeping with this idea the partial positive charge
is smeared out, \Ma mesomeric structures, over the C(8), C(10a), N(5), C(6)
and C(9a) atoms. In this generally accepted picture it is important to note that
the N(10), N(3), C(4), C(9a) and C(7) atoms are not involved in the mesomeric
structure.
Polarization of the C(2) carbonyl function in aqueous solutions gives rise
to a ir electron decrease (hence a downfield shift) at C(2) and subsequently via
mesomeric structures at C(10a), N(5), C(6), C(8) and C(9a). The downfield shift
of C(7) and C(9) shown in Figure 2A and Table 1 cannot be explained solely by
- 35 -
Figure 1: Possible mesomeric structures of oxidized flavin as deduced from NMR results.
structure b in Figure 1. Evidently structure b gives a too simplified picture
of the polarized isoalloxazine ring. To describe the polarized structure of oxi-
15 dized flavin more accurately it is necessary also to include the N chemical
shifts.
In heteroaromatic systems the nitrogen atoms can roughly be divided in
pyrrole- or a-type and pyridine- or ß-type nitrogen atoms (Witanowski et aj_.,
1972). The chemical shifts of the two classes of nitrogen atoms depend differ
ently on the polarity of the solvent. The origin of these effects lies in the
orthogonality of the electron lone pair with respect to the aromatic system in
pyridine-type nitrogens (N(l) and N(5) in flavin) and in the incorporation of
the electron lone pair into the system in pyrrole-type nitrogens (N(3) and N(10)
36
* - t
Lf>
O
S co
Lf t
• -H co
CT» CM
en
CM
o L D
co
en
Lf>
CO
CM
CO
<7>
i n
UD
o
co
CM
co
0 0
co
L D
CO
CM
O
i n
O
r^
co
CM
CO
i n
i n
Lf )
1 0
CO
co
r o
en
i-H co
KO
i n
co
c o
i e co
^ H
CO CO
o CM CO
r^
i n
co
co «st-
i n
co
U D
CO
o CM CO
O
r~-
r^
KO
co
co
CM CO
1 ^
co
«3-
i n
CM
co ro
en
co
m
CM
co
CM
KO co
r^.
co
CO
en i n
en
>d-r o
,-. CM
O
Cn i n
en
u3 co
KO
CM U 3
CM
en i n
KO
m co
co
en i n
CM
i n
o
CO
«fr KO
<3\
i n Lfï
m
i n CO
VD
o
I-N.
i n
• - H
co
i n
O
en
i n LO
f .
co
co
o VO
i n
i n
4-> t - 1 O <-H
37
C(5a)
U f a ) / C(6)
TARF (C2HCl3)
TARF (C2HCl3/MeOH)
FMN
C(4) C(2) C(10a) C(7)
LLL B
TARF (C2HCl3)
TARF (C2HCl3/MeOH)
[ ï ,3,5,10-1 \ ] MelMN
[ T 3 . 5 - 1 5 N J | FMN
160
N(5)
C(8) C(9a) C(9) C(10«)
J' I I .s IIII i' r' 150
_L _L 140
N(1)
130 120 50 45
N(3) N(10)
! )
HVJ-340 330' '200 1 9 ^ 160
PPM(ô)
150
Figure 2: Correlation diagram of 1 3C and 15N MMR chemical shifts of oxidized flavins. The 15N chemical shifts of TARF are from natural isotope abundance 15\{ NMR spectra.
in flavin). The chemical shifts of pyridine-type nitrogens show a strong upfield
shift on going from apolar to aqueous solution whereas those of the pyrrole-type
nitrogens exhibit a smaller and opposite effect. As shown in Figure 2B and Table
II the chemical shifts of the N(l), N(5) and N(3) atoms in flavin behave as ex
pected on the basis of the above given classification. The large upfield shift
of the resonances due to N(l) and N(5) observed in going from apolar solvents to
aqueous solution indicates hydrogen bond formation between these atoms and water
molecules, as already previously shown (Franken et al_., 1983). The chemical shift
of the N(3) atom shows the expected solvent dependent shift for a pyrrole-type
nitrogen. However, the chemical shift of the N(10) atom is unexpected and a dras
tic downfield shift, amounting to 15 ppm, is observed. This observation strongly
indicates that the N(10) atom is not as isolated as commonly assumed, but is
strongly affected by polarization of the molecule.
- 38 -
Table II
15 N chemical shifts of oxidized flavins as dependent on solvent and concentration.
Compound Concent ra t ion
[1 ,3 ,5- 1 5 N 3 ]FMN
FMNC)
[1 ,3 ,5 ,10- 1 5 N 4 ]MeIMN
[1 ,3 ,5 ,10- 1 5N 4 ]MeIMN
TARFC)
TARFC)
4 mM
120 mM
0.6 mM
6 mM
350 mM
350 mM
[1 ,3 ,5 ,10 - 1 5 N 4 ]Me 2 TARI e ) 7 mM
So lvent
W! 2 H 0 b > 2 H 0 b )
2 C^HC13
C2HCK/MeOHd)
rHci 3
15N Chemi
N(l)
190.8
192.9
190.5
191.1
199.9
198.8
201.1
cal s h i f t
N(3)
160.5
161.1
160.4
160.4
159.8
159.8
160.7
( in ppm)a'
N(5)
334.7
334.4
335.5
335.6
344.3
343.4
346.7
N(
-
16:
16
16
15'
15
15
a) Relative to NH, (cf. Material and Methods) b) p2H is 7.5 ó
15 c) Natural isotope abundance N NMR spectra d) Mixture contained 10% MeOH (by vol.) e) From Franken et al_. (1983).
As the N(10) atom in flavin is a substituted pyrrole-type nitrogen, not able to
form hydrogen bonds with the solvent, the large downfield shift must be ascribed to 2
a hybridization effect, i.e. in aqueous solutions N(10) possesses more sp character
than in apolar solutions. This implies that N(10) must be out of plane to some degre
in apolar solvents and more in plane in polar solvents. This configurational change
of the N(10) atom is independent of the substituent at N(10) (Table I) and is there
fore an intrinsic property of the isoalloxazine ring. It should be noted that the
chemical shift due to the C(10a) atom parallels that of the N(10) atom (Figure 2 ) .
The flat structure in polar solutions can be rationalized by the mesomeric structure
d-f in Figure 1. The creation of a partial positive charge on the N(10) atom can be
expected to lead either to an upfield shift of the resonances due to N(l), N(3),
C(4a), C(5a), C(7) and C(9) (increase of ir electron density) or to a downfield
shift of these resonances, owing to derealization of the partial positive charge.
Although these effects might influence the resonance position of N(l) and N(3) to
some degree the observed upfield shifts are caused predominantly by hydrogen bond
formation (cf. above). These two nitrogen atoms are therefore not further considered
in the following. With respect to the carbon atoms, all of them are shifted downfiel
- 39 -
indicating that the IT electron density at these sites is not increased when the
flavin is polarized and that the observed downfield shifts are caused by dereali
zation of the partial positive charge of the N(10) atom. It should be noted that
the small sensitivity of the C(4a) resonance to solvent polarity indicates that
the two opposing effects are almost cancelling each other at C(4a). It must there-
13 fore be concluded from the C NMR spectra that 0(2a) and 0(4a) are the main IT
electron acceptors upon polarization of flavin. The resulting mesomeric structures
are stabilized by hydrogen bonds to the two oxygen atoms. The mesomeric structure
d-f (Figure 1) are thus in agreement with the NMR results.
It could be argued that 0(4a) receives IT electron density from N(3). The
chemical shift of N(3) is only slightly influenced by the polarity of the solvent
(Figure 2, Table II) strongly indicating that 0(4a) receives TT electron density
mainly from the N(10) atom when the flavin is polarized. This interpretation is
in perfect agreement with coherent anti-Stokes Raman spectra of flavins (Müller
et a K , 1983). These results have shown that the frequencies of the Raman modes
are more influenced by protic solvents than by heavy atom substitution of e.g.
the N(3)H group. This effect has been ascribed to a change of bond hybridization
throughout the entire system when the flavin is polarized and hydrogen bonds are
formed. The NMR results give now detailed insights into this change of hybridi
zation.
13
In the following the effects on the C chemical shifts observed on increas
ing the polarity of the solvent will be briefly discussed. Adding methanol to
TARF in chloroform leads to a relatively strong polarization of C(2) and C(4).
The indirectly polarizable atoms, e.g. C(7), C(8), C(9) and C(10a), exhibit only
minor shifts, if any. In aqueous solutions, on the other hand, large shifts are
also evident for the indirectly polarizable atoms and C(2) and C(4) show an
additional shift. This indicates that C(2) and C(4) already polarize when some
methanol is present, i.e. when the solvent is able to form hydrogen bonds. On
the other hand, a protic solvent of a high dielectric constant is needed to pola
rize the indirectly polarizable atoms.
It has already been shown (Sarma et al., 1968; Kotowycz et al., 1969;
40
164.0-
163.5
^ 163.0
CL CL
162.5
162.0
0.01 0.1 1.0 10.0 100.0
[FMN] (mM)
Figure 3: The concentration dependence of the 13C chemical shift due to C(4) of [4-13C]FMN in aqueous solution.
Kainosho and Kyogako, 1972; Moonen and Müller, 1982a; Franken et al_., 1983) that
13 FMN in aqueous solution forms aggregates. In Figure 3 the C chemical shift of
the C(4) atom of FMN is shown in dependence of the concentration. A rather strong
13 concentration dependence of the C chemical shift is observed. Similar effects
are also observed for C(4a), C(2) and C(10a). The chemical shift values extra
polated to infinite dilution are presented in Table I. The observation that stack
ing is strongly inhibited at concentrations below 0.1 mM is in excellent agreement
13 with light absorption studies (Müller et al_. > 1973). The C NMR results also indicate that stacking of FMN inhibits its full polarization as evident from the
13 extrapolated C chemical shifts (Table I). We suggest that stacking leads to a
microenvironment with a lower effective dielectric constant than that of the pure
solvent, preventing full polarization of the molecule. It should also be noted
that the C(4) resonance is most influenced by stacking. Under this condition the
hydrogen bond to 0(4a) is weaker than in dilute solution (Table I). The concen
tration dependence of the C(4) and C(4a) resonances strongly suggests that C(4a)
becomes an important ir electron acceptor in the absence of hydrogen bonding to
0(4a) and at the same time renders 0(4a) to a less favourable -n electron acceptor.
- 41 -
This interpretation is in agreement with the fact that in monomeric FMN C(4a)
shifts downfield indicating transfer of electron density to 0(4a).
15
Similarly a comparison of the N chemical shifts as dependent on the con
centration shows that in CHC13 (Table II) the chemical shifts due to N(l), N(3)
and N(5) in TARF (natural isotope abundance spectra) are shifted somewhat upfield
as compared to those of Me„TARI at lower concentration. The latter molecule car
ries a methyl group at N(3) preventing association by intermolecular hydrogen
bond formation. In the former molecule such interactions are possible and are
reflected in a slight upfield shift of the resonances due to N(l), N(3) and N(5).
This interpretation is in full accordance with the results of TARF in CHCl,/MeOH
(Table II). More important however, is the observation that the chemical shifts
due to the nitrogen atoms in TARF and Me~TARI in CHC13 and of the corresponding
compounds in aqueous solution are very similar. This indicates that the absence
of the C(8) methyl group does not influence the particular role of the N(10) atom
on polarization. This observation is also important for our studies on flavopro-
15 teins where only MeIMN is available for N NMR studies to investigate all four
nitrogen atom of the prosthetic group.
The particular behaviour of the N(10) atom in flavin is also supported by
crystal!ographic data. Such data on oxidized flavin show that the N(10) atom
is placed somewhat out of the molecular plane (Kierkegaard et al_., 1971). In
addition theoretical calculations on the oxidized flavin molecule (Dixon et al.,
1979) have shown that only the 8.4 kJ/mol is required to bend the molecule
slightly out of the molecular plane. Furthermore, the calculated ir electron den
sities for the N(3) and the N(10) atoms are 1.731 and 1.680 electrons, respec
tively. Considering that both nitrogen atoms possess pyrrole-type character it
would be expected that the N(3) atom should resonate at higher field than the
N(10) atom. This is found for FMN in water, but not for TARF in chloroform.
Since the calculations have been performed for the coplanar structure the result
is not surprising. Consequently the incorrect prediction of the chemical shift
for the N(3) atom of TARF in chloroform can be attributed to the different struc
ture of flavin in apolar solvents.
- 42 -
The shape of the fluorescence emission and light absorption spectra and
the fluorescence quantum yield and fluorescence lifetime of flavin are strongly
dependent on the polarity of the solvent (Visser and Müller, 1979; Eweg et al.,
1979; Eweg et al_., 1980a). It is suggested that our NMR data are also related to
these phenomena. For instance the lower degree of resolution of the light absorp
tion spectrum of flavin in aqueous than in organic solvents could be related to
the existence of the mesomeric structures as presented in Figure 1.
ii§;Dihydro_flavin
In the past few years reduced flavin has attracted considerable interest,
especially because of its oxygen activating property, a reaction of great biolo
gical relevance (Massey et al_., 1969). The physical and chemical properties of
reduced flavin have been studied (Ghisla et al_., 1973; Dudley et al_., 1964).
These and crystallographic studies (Kierkegaard et al_., 1971) revealed that the
flavin molecule possesses a bent structure.
13 15 In principle a combination of C and N NMR techniques should also yield
detailed information on the solution structure of reduced flavin, as has been
demonstrated above for the oxidized molecule. In the following analysis we make
use of published data and of a personal study of several model compounds. Owing
13 to the higher complexity of the structure of reduced flavin independent C
(Van Schagen and Müller, 1980; Van Schagen and Müller, 1981) and 1 5N NMR (Franken
et al_., 1983) studies on free and protein-bound reduced flavin did not allow
development of a correlation between the chemical shifts and the structure of a
particular flavin molecule. With the data now available such a correlation appears
to exist, which is presented here. For practical reasons (sensitivity, availability
13 of compounds) C NMR results were preferably collected and supplemented by a few
15 N NMR results. The data are presented in Table III and IV. For convenience some
selected data are presented diagrammatically in Figure 4. The structure of the less
common reduced flavins are given in Figure 5.
It is evident from Table III that subsequent substitution of the various N
13 atoms of flavin influences certain C resonance lines dramatically. For conve-
- 43 -
A
5f
5e
5d
5c
5b
5a
I
I
I
I
jT.3,5,10-15N^] Me2TARIH2
[?,3,5,10-15Nt] MelMNH2
[T,3,5,10-15N^| MelMNH9
[2.3,5.-15N£| FMNH°
C(8) C<7)
C(2) C(10a)
I J
I'Y
i N i
"\C(9a),-t(6) C(5a)
ID I J
I I I
C(9)
1
C(4a) J
C(7) C(2) C(10a)C(5a)C(9a)i C(8) C(6) C(9) C(4a)
I L I I I I I I I II
) I I' 'I _L _L _L
150 KO
N(3)
130 120
N(1)
110 100
N(10) N(5)
J | i i \
i
180 160 140 120
PPM(«)
100 80 60
Figure 4: Correlation diagram of 13C and 15N NMR chemical sh i f ts of reduced f l av ins .
nience th is i s i l l u s t r a ted in Figure 4A for two sets of a few selected comparable
compounds. For instance, i t is noticeable that especial ly the quaternary atoms
C(4a), C(5a), C(9a) and C(10a) of compounds 5a-c undergo sh i f t s from 0.8 to 9.5
ppm. Yet the only difference between these compounds remains in the substituent
at N ( l ) . I t should be noticed that C(9a) and C(10a) on the one hand, and C(4a)
and C(5a), on the other hand, show para l le l s h i f t s . The C(9) atom fol lows the
sh i f t s of the l a t t e r carbon pa i r . The observed sh i f t s are c lear ly too large to be
caused solely by the e f fect of subst i tu t ion . The sh i f t s must therefore be related
to the conformation of reduced f l a v i n , indicat ing -n e lectron density changes at
the atoms in question (Van Schagen and Muller, 1980).
In case of optimal fo ld ing of reduced f l av in the benzene and the pyr imi-
5a: R1=HlR3=R5=R1o=CH3
5b: R1=R3=R1o=CH3,R5=H
5c: R1=R3=R5 = R1o=CH3
5d : R1=H,R3=CH3(Rs=COCH3lR1o=(CH2)10CH3
5e: R1=R3=R10=CH31R5=COCH3
5f : R1(R10=CH2-CH2(R3=CH3(R5=COCH3
5g:R,=R3=CH3lR5=COCH3(R1o=H
Figure 5: The structure of less common reduced flavins, used in this study (cf. Table III, IV).
dine subnuclei could be regarded as isolated ring systems. It would then be
expected that substitution of the proton at N(l) by a methyl group would only
influence the chemical shifts of the pyrimidine subnucleus. This is clearly not
observed (Figure 4A). On the other hand, substitution of the hydrogen atom at
N(3) by a methyl group (data not shown) does only slightly influence the chemical
shifts (A6 <1 ppm). This substitution does not cause severe sterical hindrance.
13 Thus the C NMR results strongly suggest that the observed shifts are caused
by a conformational change due to sterical hindrance; i.e. peri-positioned
methyl substituents at N(l) and N(10).
In the following an attempt will be undertaken to analyse the chemical shift
changes in terms of the conformation of the reduced flavin. To do this we have
- 45 -
n
E O . o .
c
i / i 4->
»*-f l/t
(O O
e - C u
o
rH
+ J e 0 )
> o l / l
c o
- p A
+-> c OJ o e o
o
—• re " O c 3 ' O O . E O
O
«o o
« er»
(_>
CJï
C_>
CO
r-*
o
uo
C J
( 0
C J
m «3-
C J
„^ ^J-
O
—*
UO
i n ^ r
CM
CO CO
'-, i n
«=*-" 1-1
m
co
CM r o
'"", co
CM CM
1—1
CM
I f ) CO
CO
CO O
L O
CO Lf>
CM
CM LO
<d"
0 0 r o
CM
co r-H
l O
m
1-1
CO
CT» CM
r o
-co 1-1
r^
co CM
1—1
r^
r^. r o
CO
r*.
r^ LO
i n
i n
r^
KO ^ r
* 3 -
r*-r o r-*
^ CT)
*"* P^.
O r o
CO
r o r o •""•
CM
CM CM
1-1
lO
** CO
co
0
CTl i n
r-*
i n
CTl
r^. «d-
CTi
<a-r o 1-1
CO
i n
* - i
^ r o
i n
CM r o
"~* 1 0
CM
'"",
O
CT* CM
CO cr>
CM
CO i n
"3-
CM u>
co
0 i n
UO
co co
** co
CT\
"""' CO
r o
CM
CO r o
^ ^ l£> CM
1—1
O
CO co
CO
i n 0
co
CO m
CM
CM i n
i n
co «d-
co
«d-r o r - .
CO
CO
"~* «* «a-r o
«a-
CM r o
'""' co
00* CM
*~' CT» «a-CM
r o
r o
CT> * 3 -
0 0 i n
-=J-
O un
^ co «d-
KO
CM CO r-i
m
m
r*
CM
r o
L O
CM c o *-H
0
r*. CM I—1
^ i n CM
r o
un 0 1
u O
r^ un
«d-
O i n
^ r*» co
CM
co" CM
"""* O
00
""H
0
en CM
UO
co CO
H
,_, 1 0
0
i n co
CM
un 0
0
r-» i n
uo
0 i n
co
«* * «3-
O co
**H
-r*.
*"• ~
d co
r o
«d-co 1-1
,_,
>d-
• d -co
co O
CM
r^. i n
^ i n
ro m ro ro ro ro co co o O 3 :
CM C J
S E
O i n
( 0 LO
o 3 :
CM O
s: E
O i n
. • un
0 n :
CM O
ï O m
0 LO
<_) 3 : CM
<_)
s : E
O i n
• 0
<_) 3 : CM
O
E
O i n
CU i n
t_> 3 :
CM t_>
S E
O un
4 -i n
<_> 3 : o 3 :
•M CM
(_)
S E
O
CT* m
0
E : E
O un
es. 3 : U . Q ;
«a:
>—
i n
3 : 0 .
E E
O CM
O
>* * r
1
1 1
1
1 1
1 1
1 1
0 0
CM O
CO
0 0 i n
.-« i n
O
i n
3 : Q .
S E
CO
0
CM 31 z s : L u
•>* l_J
ro
O
<o *3-
CM •. 3 : Z ET L u
• • *
•—'
CT» • d -i n
rt 0 co
*~H
co
'""' 0 0
0
co
CM r o 1-1
un
l u
r*.
r o r o
-0
CM
r- i n
CTi
i n
un
co
T. O .
s : E
O •d -
3 : z E Lu.
un
i n i n
1
1 1
1 1
1 1
1 1
«3-
0
r"-r^ i n
CM
co i n
\r>
co
3 : Q -
E E
r o
O
3 : z Lu.
(_) ro
0
T
• *
( N
"—'
< r
un un
crv
0
*"H
0
<£>
*~' O
CM CM
«n
UO r o r~' ,_, un
O
un co
r o
0
co
r^ i n
,_, co i n
un
cô 3 : Q .
S E
O CM
1 3 :
S
QJ E
- 46 -
Table IV
15 The N chemical shifts of reduced flavins in aqueous and chloroform solution.
Compound
[1,3,5,10-15N4]Me2TARIH2
[1,3,5,10-15N4]MeIMNH2
[1,3,5,10-15N4]MeIMNH"
[1,3,5-15N3]FMNH"
Solvent
C2HC13
pH 5.4
pH 8.5
pH 7.8
15 N chemical shift
N(l)
116.7
128.1
186.9
182.6
N(3)
145.8
150.7
150.7
149.3
(in ppm) '
N(5)
60.4
60.6
60.6
57.7
N(10)
72.2
87.3
97.2
--
a) Relative to external NH3.
to consider first the electronic structure of the N(10) and N(5) atoms which
should yield valuable information for the analysis.
If the reduced flavin were planar then both N(5) and N(10) atoms would be
pyrrole-type nitrogen atoms, i.e. both atoms are potential -n electron donors. 3
In the folded state, however, both nitrogen atoms are sp hybridized and the
electron lone pair can hardly participate in mesomeric structures. This does not
necessarily mean that both N atoms have to attain the same configuration. It is
possible that the degree of hybridization is different for both N atoms. The
degree of hybridization of the N atoms for a few available flavin derivatives can
be estimated from 1 5N NMR data (Figure 4B, Table IV). The N(l) and the N(3) atoms
resonate at lower field than the N(5) and the N(10) atoms. The chemical shifts
indicate that N(l) and N(3) are predominantly pyrrole-type, i.e. sp type, where
as N(5) and N(10) are in the region of aniline-type N atoms i.e. somewhere between
2 3 15
sp and sp type. It has already been argued (Axenrod et al_., 1971) that N che
mical shifts in aniline-type compounds reflect the participation of the nitrogen 13 electron lone pair in the TT electron system. C chemical shifts can also be used
to monitor the degree of participation of the electron lone pair of the N(5) and
N(10) atoms in the aromatic subsystems. Increasing the planarity at the N(10)
atom, for example, results in TT electron donation y_U mesomeric structures espe
cially to C(4a) and C(5a) (cf. Figure 6) leading to an upfield shift of the lat
ter atoms. On the other hand, increasing the planarity at the N(5) atom results
- 47 -
in IT electron donation to C(10a) and C(9a). The carbon pairs C(4a) and C(5a),
and C(9a) and C(10a) actually show parallel shifts, supporting our notion that
the configuration of N(10) and N(5) is mainly reflected by the chemical shift
of the mentioned carbon pairs. Cpd 5c contains methyl groups at positions 1, 3,
5 and 10 leading undoubtedly to a strong steric overlap between the methyl groups
at N(10) and N(l), and a strong overlap between the N(5) methyl group and the
carbonyl group at C(4). The downfield shifts of all four quaternary carbon atoms
in cpd 5c as compared to those of cpd 5b clearly indicate the inhibited ir elec
tron donating characters of N(5) and N(10), suggesting a strong bending of the
flavin molecule. Comparing cpd 5a with 5c it is seen that C(9a) and C(10a) show
about the same chemical shifts, whereas C(5a) and C(4a) in cpd 5a are shifted
to high field by 6.4 and 9.5 ppm, respectively. This indicates that the -n elec
tron donating character of the N(10) atom is increased in cpd 5a and that the
configuration of N(5) remains about the same in both compounds. Cpd 5b exhibits
opposite effects as cpd 5c. The ir electron donating character of N(5) is strongly
enhanced, whereas that of N(10) is decreased. These interpretations are in agree
ment with the expected steric effects in these flavins.
As evident from Figure 4A (Table III) the quaternary carbon atoms in cpd
5d-g are also \/ery sensitive to structural variations. Although the four quater
nary carbon atoms and C(9) now show almost parallel shifts, the two pairs of car
bon atoms can still be distinguished by the relative value of the shifts.
If our assumption, that the hybridization of the N(5) and N(10) atom is
reflected especially by the 1 3C chemical shifts of the C(4a), C(5a), C(9a) and
C(10a) atoms, is correct then the calculated endocyclic angles for the two
13
nitrogen atoms, using C chemical shift values, should agree with those ob
tained by X-ray crystallography. These calculations should, however, be regarded
with caution, since in these calculations we assume that only the change in the
configuration of the nitrogen centers N(5) and N(10) is contributing to the ob
served shifts of t
be fully excluded.
13 served shifts of the C resonance lines, while minor additional effects cannot
- 48 -
The endocyclic angle ip is calculated using the following equation:
6(Ci S D 3 ) " 5obs(Ci) o o
* = ate \ - *?r \ x 10-5 + 109-5 (!) 5(Ci,sp3 ) 6(Ci,sp2 )
where C. is C(4a) or C(5a) for the calculation of the endocyclic angle of N(10),
and C(9a) or C(10a) for the calculation of the endocyclic angle of N(5), <5(C. c n 3 ) 1,sp
and 6(C. 2 ) are the corresponding limit chemical shifts of carbon atom C for
a sp and sp hybridized nitrogen atom, respectively. 109.5° is the endocyclic
angle for a sp hybridized nitrogen atom and 10.5° is the difference in angles 2 3 between a sp and sp hybridized nitrogen atom. The limiting values for s(C. 3 )
i ,sp and 6(C. 2 ) m ust be known to perform the calculations. Assuming that N(10) in
2 cpd 5f approximates total sp character the endocyclic angle of N(10) in the
various compounds is calculated using the limiting values of <5(C. 2 ) = 93.3 3
ppm and 6(CC. „ 2 ) = 124.9 ppm (Table III, cpd 5f ). For an approximately sp Da, sp
hybridized N(10) atom the chemical shifts due to C(4a) and C(5a) of cpd 5c
(Figure 4A) are assumed to represent the needed limiting values, i.e. <$(C. 3) =
113.3 and 6(CC, „„3) = 141.6 ppm, respectively. The endocyclic angle of the N(10) oa ,sp 13 atom of various flavin derivatives, as determined from experimental C chemical
shift values, are presented in Table V, together with published crystallographic
data. Table V demonstrates that the two sets of data agree surprisingly well,
indicating that our method of determining the endocyclic angle of the N(10) atom
in reduced flavin can be used reasonably safe. The minor differences between the
two sets of data are probably caused by the fact that we have completely neglected
possible effects on the chemical shifts of the atoms in question by substitu
tion of nearby atoms. Apparently such effects are only of minor importance.
The determination of the endocyclic angle of the N(5) atom is much more
difficult because no safe indications are available demonstrating that a partic-
ular 1,5-dihydroflavin possesses a sp hybridized N(5) atom. We therefore, used
the following modified procedure. It is known from crystallographic studies
(Norrestam and Von Glenn, 1972) that the N(5) atom in cpd 5b possesses an endo
cyclic angle of 116.4°, i.e. the atom is about 65% sp and 35% sp hybridized.
- 49 -
Table V Comparison of the endocyclic angles of the N(10) and N(5) atoms in reduced flavins as determined
by crystallographic and NMR methods. For details, see text.
Compound Endocyclic Angles of Nitrogen Atom
N(10) N(5)
X"ray 6C(4a) óC(5a) X"ray äC(10a) 6C(9a)
5a -- 114.5° 113.5° -- 112.6° 112.0° a) a) . \ .*
5b 111.9° 110.3° 111.9° 116.4° -- °> - b>
5c -- - D) -- D) -- 112.0° 112.5° c) c)
5d 116.8° 117.5° 117.4° 113.8° 111.4° 114.1° d) d)
5e 113.0° 113.4° 114.9° 112.0° 109.9° 111.7°
5f -- -- b) -- b) -- 111.1° 114.4° e) e)
5g 117.0° 118.9° 119.5° 114.8° 111.3° 115.5°
TARFH, -- 113.8° 113.0° - 117.1° 118.3° J) f) f) f)
FMNH2 - 114.9° 114.4° -- 113.3° 116.9°
a) Taken from Norrestam and von Glehn (1972).
b) Values used in the calculations, see text.
c) Taken from Norrestam et ah (1969).
d) Taken from Werner and Rb'nnquist (1970).
e) Taken from Leijonmarck and Werner (1971).
f ) These values must be.considered with great caution, for an explanation, see text.
1 15 1 The degree of hybridization is in agreement with the J( N- H) coupling constant
of 87.5 Hz for a similar compound (Franken et a]_., 1983). Thus the chemical shift
values of the C(9a) and C(10a) atoms of cpd 5b (Table III) are used as reference
chemical shifts for an endocyclic angle of 116.4° (Eq. 1). In addition it was as-
sumed that a change from sp to sp hybridization of N(5) causes similar chemical
shift changes of C(9a) and C(10a), as a hybridization change of N(10) on the chem
ical shift of C(4a) and C(5a), as discussed above. Although somewhat arbitrary,
this assumption is supported by the fact that the introduction of peri-over
crowding effects between a N(5) substituent and the C(4) carbonyl function causes
chemical shift changes at the C(9a) and C(10a) centers comparable with analogous
effects on the chemical shift changes at the C(4a) and the C(5a) centers by N(l)
and N(10) substituents. In addition the value of 109.5° in Eq. 1 was replaced
by 116.4° for the calculations.
The calculated values are compared with crystallographic data in Table V.
- 50 -
The endocyclic angles calculated from NMR results are in fair agreement with the
crystallographic data. The good agreement supports the validity of our empirical
approach, both for the N(5) and the N(10) atom. Therefore, the data provide a
basis for the calculation of endocyclic angles of compounds which cannot be stud
ied by crystallographic methods. We wish to emphasize that our calculations of
the endocyclic angles of the N(5) and the N(10) atoms do not yield numerical va
lues on the degree of folding of the reduced flavin molecule, although the endo
cyclic angles of the two atoms are related to the degree of folding of the molec
ule. Since the degree of hybridization of the two nitrogen atoms can be independ
ently modulated, a fact also supported by crystallographic studies (Norrestam et
al., 1969; Werner and Rönnquist, 1970; Leijonmarck and Werner, 1971; Norrestam
and Von Glehn, 1972), the term "folding angle of the flavin molecule" should be
avoided. In fact the degree of hybridization of the nitrogen atoms describes the
structure of the reduced flavin molecule more accurately than the angle of folding.
15 Furthermore, it should be mentioned that N chemical shifts of N(5) acetylated
compounds cannot be used to calculate the N(5) endocyclic angle since this sub-
15 stitution will undoubtedly influence the N chemical shift of the N(5) atom.
In Table V the calculated endocyclic angles for TARFH? are also given. The 2
results indicate that the N(5) atom is considerably sp hybridized. This result
1 15 1 is in excellent agreement with the J( N(5)- H) coupling constant of TARFH2
2 (Franken et al_., 1983). The N(10) atom, on the other hand, possesses less sp
character than the N(5) atom. The results indicate that TARFH,, possesses a fairly
planar conformation, more bent at the N(10) than the N(5) center.
The calculated endocyclic angles for FMNhL in aqueous solution are also
presented in Table V. The calculated endocyclic angle for the N(5) atom, espe-
13 cially that calculated from the chemical shift of C(10a), is smaller than
that for TARFH?. Based on published light absorption data (Dudley et al_., 1964)
it was expected that the endocyclic angle of the N(5) atom in FMNH,, would be at
least as large as that in TARFH?. The reason for this apparent discrepancy will
13 15 be explained in the following on the basis of the C and N chemical shifts
of FMNH2 (Table III and IV). As compared to Me2TARIH2 in CHC13 the 1 5N chemical
shifts of MeIMNH2 in aqueous solution appear, with the exception of that of the
- 51 -
N(5) atom, at lower fields (Figure 4B, Table IV). The large downfield shift of 2
the resonance line of N(10) indicates a drastic change to enhanced sp character. ?
The increased sp character of N(10) is also reflected in the upfield shift of
the resonances due to C(4a) and C(5a). This upfield shift is however less than
would be expected on the basis of the downfield shift of the N(10) resonance.
The N(5)H and N(3)H groups in reduced flavin form hydrogen bonds with the solvent
15 in aqueous solution leading to downfield shifts of the corresponding N chemical
shifts. This is clearly observed on the chemical shift of N(3). The small effect
observed on the N(5) resonance suggests that the expected downfield shift is al-
most cancelled by an increase of the sp character of N(5). This interpretation
is supported by the downfield shift of the resonances due to C(9a) and C(10a),
but the downfield shift observed for C(10a) is much larger than expected by the
change of hybridization of the N(5) atom. Changing the solvent from CHC1, to
aqueous solution also leads to an unexpectedly large downfield shift of the re
sonance line of N(l). This downfield shift of the resonance of the pyrrole-type
N(l) atom is too large to originate solely from hydrogen bond formation. The large
downfield shift suggests a considerable increase in sp character of the N(l)
atom. The similar (in magnitude) and parallel shifts of the resonances due to the
N(l) and N(10) atoms indicate that both atoms can in some way act cooperatively.
In contrast to the oxidized flavin the chemical shifts due to C(2) and C(4) are
only slightly affected when going from CHC1, to aqueous solution. This indicates
that the change in solvent hardly increases the partial positive charge of the
C(2) and the C(4) centers. The explanation is that the IT electron density of the
0(2o) and 0(4o) centers is increased by a cooperative electron donation from
N(10) and N(l), and even from N(3). As a consequence 0(2a) and 0(4a) can no long
er polarize C(2) and C(4) to such an extent as in oxidized flavin. It should
13 be noted that the C chemical shifts are only slightly dependent on concentration, although aggregation does occur as judged from the increase of linewidths
upon increasing the concentration.
13 In our opinion the electronic structure of FMNH« can, based on the C
15 and N NMR data, best be described by the structures shown in Figure 6. The
- 52 -
Figure 6: Possible mesomeric structures of reduced flavin as deduced from NMR results.
existance of the tautomeric structures b and c cannot be directly deduced from
the NMR results but we believe that they are important to understand the elec
tronic structure of reduced flavin. Structures e and f are in full agreement
with the parallel large downfield shifts of the resonances due to N(l) and N(10)
and also explain the unexpected large downfield shift of the resonance of C(10a).
The latter observation is also in agreement with structure d. All three mesomeric
structures are in accord with the observation that the IT electron density at the
C(4a) center is increased in FMNH2, as compared to TARFH„. On the other hand the
3 15
N(5) atom is forced to acquire more sp character, as deduced from N NMR data.
This implies that IT electron density will be withdrawn from the centers C(10a),
C(9a), C(6) and C(8). The downfield shifts of these atoms support this idea.
It becomes evident from the discussion above that our semi-empirical approach
to calculate the endocyclic angles of reduced flavin in aqueous solution only
yields lower limits. In our calculations we assumed that the change in hybridiza
tion of the N(5) and N(10) atoms will lead to a u electron density increase or
decrease at the centers C(9a) and C(10a), and C(4a) and C(5a), respectively. In
apolar solutions where mesomeric structures, as presented in Figure 6, are not
favoured, the semi-empirical approach yields reasonably good results. In highly
- 53 -15
polar solvents, where mesomeric structures play a role, the N chemical shifts
due to the N(5) and N(10) atoms are more reliable parameters to estimate the
13 endocyclic angles. At any rate it is estimated from the combination of C and
15N NMR results that in FMNH2 the endocyclic angle for N(10) is about 116-118°
and that for N(5) is slightly less.
The N(1)H group of reduced flavin deprotonates with a pK of 6.6 (Dudley
et aj_., 1964; Van Schagen and Müller, 1981; Franken et aj_., 1983). The negative
15 charge at N(l) leads to a strong downfield shift of the N chemical shift of
13 N(l) and the C chemical shifts of the neighbouring carbon atoms C(2) and C(10a).
These shifts are similar to those observed in other heteroaromatic systems (Pug-
mire and Grant, 1968,1971; Ewers et a]_., 1974; Tokuhiro and Frankel, 1969). The
13 13
C(2) resonance is shifted less downfield than the C(10a) resonance, owing to
the increased -n electron density at 0(2o). It was expected that the negative
charge at N(l) would increase the IT electron density at N(10), leading to an up-
field shift of the resonance of N(10). In contrast a downfield shift of 10 ppm 2 2
is observed, indicating an increase of sp character of N(10). The increased sp
character of N(10) is also reflected in the small upfield shift of the resonances
due to C(4a), C(5a) and C(7). Deprotonation of reduced flavin influences the chem
ical shift of the N(5) resonance only slightly, the small upfield shift indicates
a slight decrease of the predominant sp character of the N(5) atom. These re
sults indicate that the N(10) atom in ionized reduced flavin possesses almost
full sp character, while that of the N(5) atoms is somewhat less.
The analysis presented above shows that the conformation of reduced flavin
is mainly governed by the polarity of the solvent and by steric hindrance intro
duced by N-substitution. In polar solvents, where mesomeric structures are sta
bilized, unsubstituted reduced flavin attains an almost planar structure. It
has been suggested that the pyrazine subnucleus of reduced flavin possesses anti-
aromatic character causing a strong bending of the molecule (Tauscher et al.,
1973). The mesomeric structures presented in Figure 6 show that the molecule
can 'escape' this anti-aromaticity and attain a fairly coplanar structure. The
previous interpretation of the light absorption spectra of reduced flavin (Dudley
et al^., 1964) should be reconsidered in view of our NMR results, i.e. the observed
- 54 -
molar absorption coefficient at 450 nm for a particular flavin is most probably 2
solely related to the degree of sp hybridization of the N(5) atom of reduced
flavin. Our results are also in agreement with theoretical calculations by Dixon
et al_. (1979), who showed that the reduced flavin molecule is only slightly bent
and that only a relatively small activation energy is needed for the molecule to
acquire a planar conformation. These results were recently supported by photo-
electron spectroscopy (Eweg et al., 1980b).
Our results have several biological implications. It has been shown by
crystallographic studies (Ludwig et aj_., 1976) that FMNH" is almost coplanar when
bound to Clostridium MP flavodoxin. In M.elsdenii flavodoxin which is closely
related to Clostridium MP flavodoxin, the sp character of the N(5) atom of
the reduced prosthetic group is confirmed by the coupling constant (Franken
et al_., 1983). Simondsen and Tollin (1980) suggested that the planar structure
of FMNH" in these proteins is mainly responsible for the altered redox potential
of the couple semi qui none/fully reduced, as compared to free FMNH". Our results,
however, indicate that free FMNH" already possesses an almost planar conforma
tion. It is therefore unlikely that the redox potential in flavodoxin is governed
mainly by the conformation of protein-bound FMNH". Probably other factors such as
negative charges in the vicinity of the bound flavin play a much more important
role.
In this paper we have presented a detailed analysis for the interpretation
of the chemical shifts of free flavin which provides a good basis for NMR studies
on flavoproteins. The easy manner in which the structure of reduced flavin is
altered by the introduction of sterical hindrance or hydrogen bonds is striking.
It is suggested that these effects are involved in directing the different path
ways of flavin catalysis. Reduced flavin reacts with molecular oxygen at the
C(4a) position, as was deduced by Ghisla et al_. (1978) from NMR data on a luci-
ferase. This reaction will be favoured if the TT electron density at the C(4a) 9
atom is enhanced. As we have seen this can be accomplished by sp hybridization
of the N(10) atom, but also by providing a hydrophobic-!ike environment where
mesomeric structures are not favoured. These aspects are currently under study
in our laboratory.
- 55 -
ACKNOWLEDGEMENTS
We are indebted to Mrs. J.C. Toppenberg-Fang for typing the manuscript,
to Mr. M.M. Bouwmans for preparing the figures and to Dr. C.A.H. Rasmussen for
carefully reading the manuscript.
REFERENCES
Axenrod, T., Bregosin, P.S., Wieder, M.J., Becker, G.D., Bradley, R.B. and Milne, G.W.A. (1971) J.Am.Chem.Soc. 93, 6536-6541.
Dixon, D.A., Lindner, D.L., Branchard, B. and Lipscomb, W.N. (1979) Biochemistry 18, 5770-5775.
DudTey, K.H., Ehrenberg, A., Hemmerich, P. and Müller, F. (1964) Helv.Chim.Acta 47, 1354-1383.
Eweg, J.K., Müller, F., Visser, A.J.W.G., Veeger, C , Bebelaar, D. and van Voorst, J.D.W. (1979) Photochem.Photobiol. 30, 463-471.
Eweg, J.K., Müller, F., Bebelaar, D. and Van Voorst, J.D.W. (1980a) Photochem. Photobiol. 31, 435-443.
Eweg, J.K., Müller, F., Van Dam, H., Terpstra, A. and Oskam, A. (1980b) J.Am. Chem.Soc. 102, 51-61.
Ewers, U., Günther, H. and Jaenicke, L. (1974) Chem.Ber. 107, 876-886. Franken, H.-D., Rüterjans, H. and Müller, F. (1983), E"ür.J.Biochem., submitted. Ghisla, S., Hartmann, U., Hemmerich, P. and Müller, F". (1973) Liebigs Ann.Chem.,
1388-1415. Ghisla, S., Hastings, J.W., Favaudon, V. and Lhoste, J.M. (1978) Proc.Natl.Acad.
Sei.USA 75, 5860-5863. Grande, H.TT, Gast, R., Van Schagen, C G . , Van Berkel, W.J.H. and Müller, F.
(1977a) Helv.Chim.Acta 60, 367-379. Grande, H.J., Van Schagen, C G . , Jarbandhan, T. and Müller, F., (1977b) Helv.
Chim.Acta 60, 348-366. Hemmerich, P. and Massey, V. (1982) in: Oxidases and Related Redox Systems
(King, T.E., Mason, H.S., Morrison, M., eds.) Pergamon, New York, pp. 379-405. Kainosho, M. and Kyogoka, Y. (1972) Biochemistry 11, 741-752. Karplus, M. and Pople, J.A. (1963) J.Chem.Phys. 38T 2803-2808. Kawano, K., Ohishi, N., Suzuki, A.T., Kyogoku, Y. and Yagi, K. (1978) Biochemistry
17, 3854-3859. Kierkegaard, P., Norrestam, R., Werner, P.-E., Csöregh, I., Von Glehn, M.,
Karlsson, R., Leijonmark, M., Rönnquist, 0., Stensland, B., Tillberg, 0. and Torbjörnssen, C (1971) in: Flavins and Flavoproteins (Kamin, H., ed.) University Park Press, Baltimore, pp. 1-22.
Kotowycz, G., Teng, N., Klein, P. and Calvin, M. (1969) J.Biol.Chem. 244, 5656-5662.
Lauterbur, P.C. (1962) J.Am.Chem.Soc. 83, 1846-Lauterbur, P.C. (1965) J.Chem.Phys7~4"3, 360-363. Leijonmarck, M.and Werner, P.-E. (1971) Acta Chem.Scand. 25, 2273-2290. Levy, G.C. and Lichter, R.L. (1979) in: Nitrogen-15 Nuclear Magnetic Resonance
Spectroscopy, John Wiley and Sons, New York, pp. 28-107. Ludwig, M L . , Burnett, R.M., Darling, G.D., Jordan, S.R., Kendall, D.S. and
Smith, W.W. (1976) in: Flavins and Flavoproteins (Singer, T.P., Ed.) Elsevier, Amsterdam, pp. 393-404.
Massey, V., Müller, F., Feldberg, R., Schuman, M., Sullivan, P.A., Howell, L.G., Mayhew, S.G., Mathews, R.G. and Foust, G.P. (1969), J.Biol.Chem., 244, 3999-4006.
Moonen, C.T.W, and Müller, F. (1982a) Biochemistry 21, 408-414.
56 -
Müller, F., Kaptein, R. and Mayhew, S.G. (1982b) Moonen, C.T.W., Höre, P.J. FEBS Lett. 149, 141-146.
Moonen, C.T.W, and Müller, F. (1983) Eur.J.Biochem. 133, 463-470. Müller, F., Hemmerich, P., Ehrenberg, A., Palmer, G. and Massey, V.
Eur.J.Biochem. 14, 185-196. Müller, F. (1971)Tn: Methods Enzymology 18B, 453-458. Müller, F. (1972) Z.Naturforsch. 27b, 102CT026.
(1970)
Mayhew, S.G. and Massey, V. (1973), Biochemistry, 12, 4654-4662. Horowitz, M. and Carreira, UTA. (1983)
and Torbjörnsson, L. (1969)
Müller, F. Müller, F., Vervoort, J., Lee, J.
J.Raman Spectrosc. 14, 106-117. Norrestam, R., Kierkegaard, P., Stensland, B.
Chem.Comm., 1250-1251. and Von Glenn, M. (1972) Acta Cryst. B28, 434-440. and Grant, D.M. (1968) J.Am.Chem.Soc.~9"0", 697-706. and Grant, D.M. (1971) J.Am.Chem.Soc. 93, 1880-1887.
Lyerla, J.R., Peat, I.R., Cohen, J.S., Reynolds, W.F. 96, 570-574.
Norrestam, R. Pugmire, R.J. Pugmire, R.J. Quirt, A.R.
M.H. (1974) J.Am.Chem.Soc. and Freedman,
Sarma, R.H., Danner, P. and Kaplan, N.O. (1968) Biochemistry 7, 4359-4367. Scola-Nagelschneider, G. and Hemmerich, P. (1976) Eur.J.Biochem. 66, 567-577. Simondsen, R.P. and Tollin, G. (1980) Mol.Cell.Biochem. 33, 13-24. Tauscher, L., Ghisla, S. and Hemmerich, P. (1973) Helv.CïïTm.Acta 56, 630-644. Tokuhiro, T. and Frankel, G. (1969) J.Am.Chem.Soc. 91, 5005-5013. Van Schagen, C G . and Müller, F. (1980) Helv.Chim.Acta 63, 2187-2201. Van Schagen, C G . and Müller, F. (1981) Eur.J.Biochem. 170, 33-39. Visser, A.J.W.G. and Müller, F. Werner, P.-E. and Rönnquist, 0. Witanowski, M., Stefaniak,
Tetrahedron 28, 637-653. Witanowski, M., Stefaniak,
(1979) H'elv.Chim.Acta 6FT593-608. (1970) Acta Chem.Scänd. 24, 997-1009.
L., Januszewski, H. and Grabowski, Z. (1972)
and Webb, G.A. (1981) in: Annual Report Nuclear Magnetic Resonance Spectroscopy (Webb, G.A., Ed.) Vol. IIB, pp. 1-493.
Yagi, K., Ohishi, N., Takai, A., Kawano, K. and Kyogoka, Y. (1976) Biochemistry 15, 2877-2880.
- 57 -
Chapter 6
A CARBON-13 NUCLEAR MAGNETIC RESONANCE STUDY ON THE DYNAMICS OF THE CONFORMATION
OF REDUCED FLAVIN
Chrit T.W. Moonen, Jacques Vervoort and Franz Müller
- 58 -
SUMMARY
Several flavin model compounds in the reduced state have been investigated
13 by C NMR techniques. The NMR spectra were recorded in dependence of temperature,
in the range of 30° to -100°C. The results show that the interpretation of similar
results by Tauscher et al_. (1973) are incorrect. Nitrogen inversion, as previously
suggested by these authors, is not observed, but ring inversion is. The ring in
version is strongly dependent on solvent polarity and hybridization of the N(5)
atom. Symmetry arguments in the dynamic processes of the inversion reactions are
discussed.
INTRODUCTION
Reduced (1,5-dihydro)flavin plays an important role in biological reactions
catalyzed by flavoproteins. It is known from crystallographic studies (Kierkegaard
et a^., 1971) on flavin model compounds that reduced flavin possesses a bent con
formation, often referred to as "butterfly" conformation. It has been proposed
that the activation barrier for the transition from the bent to the planar con
formation provides a means of regulating the redox potential of protein-bound
flavin (Simondsen and Tollin, 1980; Tauscher et al_., 1973; Van Schagen and Müller,
1981). In this concept the apoflavoprotein forces the bound flavin to attain a
certain conformation. This concept seems to be supported by theoretical calcula
tions indicating that an activation barrier of 145 kJ/mol exists between the bent
and the planar conformation (Palmer and Platenkamp, 1979). Tauscher et a K (1973)
investigated the nitrogen inversion in reduced flavin by H NMR. An activation
barrier of about 42 kJ/mol for the inversion of the N(5) atom was calculated from
13 15 these results. This seems to be in contradiction with our recent C and N NMR
data (Moonen et al_., 1984). These results demonstrate that the conformation of
reduced flavin is mainly governed by intramolecular overcrowding effects (substi
tution at N(l) and N(5)) and the polarity of the solvent. The ease in which the
endocyclic angle of the N(5) and the N(10) atom can be modulated independently is
especially striking. This suggests that the activation barrier for reduced flavin
- 59 -
to attain a planar conformation must be low.
13 Preliminary C NMR results (van Schagen and Muller, 1980) strongly indi
cated that this technique is particularly useful for investigation of the dynamics
13 of the conformational changes of reduced flavin. Since the C chemical shifts
of flavin are mainly governed by the IT electron density at the carbon atoms under
13 study (Van Schagen and Müller, 1980), C NMR should yield a much better insight
into the dynamics of the conformational change than H NMR where only a limited
13 number of nuclei in flavin are available. Furthermore the resolution of C
spectra at high field strength is much better than the splitting of a single
proton resonance line to an AB pattern, as used by Tauscher et al_. (1973). Our
13 results demonstrate that the temperature dependent C NMR spectra of reduced
flavin yield detailed information on the conformation of reduced flavin and
allow calculation of a more accurate value for the activation barrier. This
study also shows that nitrogen inversion is not observed and that the spectral
changes are related to the "butterfly" motion of the molecule.
This is in contrast to published results (Tauscher et al_., 1973).
MATERIALS AND METHODS
The compounds used in this study were prepared according to published pro
cedures: l,3,7,8,10-pentamethyl-5-acetyl-l,5-dihydro-isoalloxazine, 1,3,7,8,10-
pentamethyl-5-benzyl-l,5-dihydro-isoalloxazine, 1,3,5,7,8,10-hexamethyl-1,5-
di hydro-isoal1oxazi ne and 1,3,7,8-tetramethyl-5-acety1-5,10-di hydroal1oxazi ne,
(Dudley and Hemmerich, 1967); 3,7,8-trimethyl-l,10-ethano-5-acetyl-l,5-dihydro-
isoalloxazine (Müller and Massey, 1969), the parent compound was prepared from
riboflavin (Grande et al_., 1977); 3-methyl-5-acetyl-l,5-dihydro-tetraacetylribo-
flavin (van Schagen and Müller, 1980); 3-methyl-l,5-dihydro-tetraacetylriboflavin
was prepared directly in the NMR tube by a two-phase reduction (van Schagen and
Muller, 1980). Subsequently the aqueous phase was removed by means of a syringe
under anaerobic conditions.
13 Wilmad 10 mm precision tubes were used. C NMR measurements were performed
at 75.5 MHz on a Bruker 300 CXP spectrometer. Broadband decoupling (3 Watt) was
- 60 -
applied. A repetition time of 1 s was used. Pulse width was 30 . 5000 transients
were usually accumulated. Conversion rates were determined from the coalescence
points according to the method of Binsch (1975). Spectra exhibiting unequal popu
lations of the various forms were analyzed using the method of Shanan-Atidi and
Bar-Eli (1970). Activation barriers were determined using the Eyring equation as
described by Binsch (1975).
RESULTS AND DISCUSSION
As has been discussed previously by Tauscher et al_. (1973) the following
dynamic processes have to be considered by the analysis of temperature-dependent
conformations of reduced flavin: 1) the rotation of the substituent at N(5)
around the N(5) - N(5a) bond; 2) the inversion of the N(5) and the N(10)atoms;
and 3) the "butterfly" motion of the reduced flavin molecule, i.e. the flipping
around the N(5)-N(10) axis. It is often difficult to discriminate experimentally
between nitrogen inversion and ring inversion ( "butterfly" motion) in hetero-
cycles (Lambert et a]_., 1971). Some certainty with regard to this problem can
be obtained using various model compounds, as will be shown below.
The structure of the compounds investigated in this study are shown in
13 Figure 1. The dependence on the temperature of the C NMR-spectra of cpds. 1
and 2 is shown in Figures 2 and 3, respectively. At low temperature both com
pounds give rise to splitting of all or almost all resonances, although the
population stoichiometry of the splitting is different. The activation barrier
13 calculated from C NMR spectra for several compounds is summarized in Table 1.
The question now arises which dynamic process causes the observed splitting of
the resonances. Solely freezing the rotation of the acetyl group would probably
not cause splitting of the resonances due to the carbon atoms of flavin. At most
splitting of the resonances due to the acetyl side chain would be expected. The
splitting patterns shown in Figures 2 and 3, therefore, reflect a dynamic pro
cess, which influences the electronic structure of the whole isoalloxazine ring.
This effect could be caused by inversion of one of the nitrogen atoms (N(5) or
61
N — C H 3
1: R>R1o=CH2-CH2lR5=COCH3 2: R1=R1o=CH3,R5=COCH3
3: R1=CH3lR1o=H,R5=COCH3
l : R1=H(R10=tetraactylribity[,R5=COCH3
5: R1=R5=H,R10=tetraactylribityl 6: R^=R^Q=CH3 R5=CyHy 7: Ri=R5=R10=CH3
Figure 1: The structure of the compounds investigated.
N(10)) or by the so-called butterfly motion or by a combination of effects,
both resulting in a change of the nitrogen substituents from axial to equatorial
position and vice versa. An inversion of the N(10) atom can be excluded since the
gross effects observed for cpds. 1 and 2 are \/ery similar, with almost equal
activation barriers. As has been shown by Moonen et a]_. (1984), cpd. 1 contains
an sp hydridized N(10) atom, whereas the N(10) atom in cpd. 2 exhibits consider-o
able sp character. Nitrogen inversion is strongly dependent on the endocyclic
angle and consequently on the hybridization of the nitrogen atom. (Lehn and
Wagner, 1970; Lambert et al_., 1971).
Although the hybridization of the N(10) atom differs in cpds. 1 and 2, very
similar effects are observed for both compounds. We therefore conclude that
the splitting patterns are not caused by N(10) inversion. This leaves us with
the dynamic process of the N(5) inversion and/or the "butterfly" motion. Which
ever process predominates, a partial rotation of the acetyl group is needed, as
demonstrated in Figure 4. The acetyl function has to align along the N(5)-N(10)
axis before ring inversion or N(5) inversion can take place. Due to considerable
overlap between the carbonyl function at C(4) and the acetyl group at N(5),
- 62
33 C
_JL_
2°C
-15 C
-30 C
-60 C
JL
JUL
J LL
J_™___jJ_
J L L Lx
JJLLJLJL.
JL
l I .J l i l . J L x l J _
H9C — CH 9 2 I I
JL.
JL
X .
170 150 130
PPM (a)
110 90
Figure 2: Temperature dependence o f the na tu ra l abundance 13C NMR spectrum of 3 ,7,8-tr imethyl- l ,10-ethano-5-acetyl- l ,5-dihydroisoal loxazine (1) in chloroform.
th is leads to an energet ical ly rather unfavorable s i t ua t i on . This also means
that the act ivat ion barr ier of the two possible processes i s strongly influenced
by the act ivat ion barr ier of the acetyl r o ta t ion . The ac t ivat ion bar r ier of the
acetyl ro tat ion can be roughly estimated from published data on amino and amide
rotat ion ( for a review, see Binsch (1975)). A ro tat ion barr ier of 39.5 kJ/mol
was determined for example, for l,l,l,-trimethyl-2,2-dichloroaminomethane, con-3
ta in ing an sp hybridized nitrogen atom (Bushweiler et al_., 1973). In amides,
possessing a par t ia l double bond, the barr ier was found to be about 60 kJ/mol
i f only a small s te r ic hindrance is present in the molecule, and the barriers
- 63 -
33"C
J UL JL
2°C
-15 C
CH, CH, I 3 I :xox3
A H3C^C^0°
-30°C
-60°C
_L
A
170 150 130
PPM («)
110
Fiaure 3- Temperature dependence of the natural abundance 13C NMR spectrum of ' i ,3,7,8,10-pentamethyl-5-acetyl-l,5-dihydroisoalloxazine (2) in
chloroform.
become much higher i f s te r ic hindrance is increased (Binsch, 1975).
The activation barrier for the acetyl rotation in flavin is estimated from
13 these data to be a t least 60 kJ/mol. Furthermore, C NMR spectra of cpd. 7 in
13 chloroform does not show a sp l i t t ing or even a broadening of the C l ines a t
-60 C which clearly demonstrates that the actual activation barrier for the ring
inversion or N(5) inversion must be rather low. (data not shown). Tauscher e t a l .
- 64 -
Table 1 Activation barriers in chloroform calculated from the temperature
13 dependent C NMR spectra using the method of Shanan-Otidi and Bar-Eli
(1970)
Compound Activation barrier
AG (kJ/mol) at the
alescence temperature
55.2
55.6
57.6
61.9
Coalescence
temperature
(°K)
258
273
273
306
(1973) suggested that ring inversion in flavin is energetically the most favored
process. This suggestion is in agreement with published data obtained on different
compounds. For instance, Lehn and Wagner (1970) and Lambert et al_. (1971) showed
that the activation barrier for nitrogen inversion in cyclic compounds ranges
3
from 40 kJ/mol to much higher values for sp hybridized nitrogen atoms not con
taining a covalently bound hydrogen atom. On the other hand, ring inversion in
cyclohexane shows an activation barrier of 42 kJ/mol (Anet and Anet, 1975). In
troduction of a double bond into this system however, markedly decreases the
activation barrier (Anet and Anet, 1975). Bernard and St. Jacques (1973) deter
mined the barrier in cyclohexene to be 20-25 kJ/mol. The drastically decreased
barrier was ascribed to the fact that cyclohexene cannot adopt the low energy
chair conformation. This means that the more planar the conformation of a molec
ule becomes the lower the barrier will be, as has also been demonstrated for 2
piperidine and derivatives thereof, containing a sp center (Gerig, 1968;
Jensen and Beck, 1968; Lambert and Oliver, 1968). The barriers determined for
cyclohexane, cyclohexene and the piperidine derivatives are of about the same
magnitude. This also strongly indicates that the replacement of a carbon atom
in cyclohexane by a nitrogen atom does not influence the activation barrier
much. This suggests that the properties of the pyrazine subnucleus of reduced
65
B
•: o ;
Figure 4: Schematic representation of the "butterfly" motion and the rotation of the N(5) acetyl side chain. The configuration of the acetyl side chain in A is favoured as the steric overcrowding effect with the carbonyl group at position 4 is minimized (also the para-overcrowding effect with the N(10) substituent is minimized). To allow the "butterfly" motion (B -> C) it is necessary that the acetyl side chain rotates (A -» B) to minimize the peri-overcrowding effect with the C(4) carbonyl group during the planar transition state, (see also text).
flavin are compatible with those of simple cyclohexane and cyclohexene deriva
tives. Thus, the actual activation barrier for ring inversion must indeed be
far below 60 kJ/mol and the apparent high activation barriers reported in this
study are caused by the high barrier for the acetyl rotation.
From this it can also be concluded that ring inversion in reduced flavin is
considerably more favored than inversion of the acetylated N(5) atom. This im
plies that we are observing the effects of the ring inversion in Figures 2 and 3,
although the calculated activation barriers from the splitting patterns repre
sent its activation energy increased by the barrier of the unfavorable acetyl
rotation.
The splitting patterns of all N(5) acetylated flavins investigated are rather
similar and the activation barriers hardly differ (Table 1). This indicates that
the activation barrier for the ring inversion is hardly dependent on the N(10)
substituent (see Table 1). As the N(10) atom in flavin is always more sp hybri
dized and steric effects are less for the N(10) than for the N(5) atom, it is
- 66 -
not surprising that the activation energy for the overall dynamic process is still
mainly determined by the acetyl rotation at the N(5) atom. In MeTARFH„, on the
other hand, no splitting at all could be observed down to -40°C. In this compound
the N(5) atom contains a hydrogen atom and therefore the butterfly motion is greatly
facilitated, in accordance with our results.
This interpretation is in full agreement with earlier observations (Moonen et al.,
1984) showing that the degree of hybridization of the N(5) and the N(10) atoms in
reduced flavin can easily be influenced by substitution. Dixon et al_. (1979) cal
culated an activation barrier of 16 kJ/mol for the "butterfly" motion of unsub-
stituted 1,5-dihydroflavin. In our opinion this value is, also considering the
published results on cyclohexane, cyclohexene and pi peri dine derivatives, close
to the actual activation barrier. This is also in agreement with our results on
TARFH~. In contrast, the theoretical value of 145 kJ/mol for the ring inversion
of reduced flavin, calculated by Palmer and Platenkamp(1979), must be incorrect.
The interpretation of our results is in contradiction to work published by
Tauscher et al_. (1973). These authors investigated H NMR spectra of N(5) alky
lated flavin derivatives in dependence of temperature and ascribed the observed
effects to the inversion of the N(5) atom. They excluded a rotation of the N(5)
side chain around the N(5)-C(5a)- or C(5a)-C(5ß)-axis. This conclusion was based
on the fact that l,3,7,8-tetramethyl-10-benzyl-5-acetyl-l,5-dihydro-isoalloxazine
did not show a splitting of the methyl group of the N(5) acetyl group, not even
at -107 C. It is, however, by no means certain that the two possible rotamers
can be resolved using H NMR. Even if the two rotamers could be resolved, their
detection could well be obscured by a large population difference. Furthermore,
the maximum activation energy for the acetyl rotation, derived by Tauscher et
al. (1973) from the absence of a splitting in the H NMR spectrum of 1,3,7,8-
tetramethyl-10-benzyl-5-acetyl-l,5-dihydroisoalloxazine is far below what would
be expected for such a rotation (cf. Binsch, 1975), taking into consideration
a) the partial double bond character in the amide-like N(5)-C(5a) bond,
b) the peri-overcrowding with the carbonyl function at C(4) and
c) the para-overcrowding with the large N(10) benzyl substituent.
- 67 -
Moreover, it can be concluded directly from the published spectra of Tauscher
et al_. (1973) that the interpretation of their data is incorrect. If it were
true that rotation of the N(5) side chain in N(5) benzylated flavin is fast on
the NMR timescale down to -100°C, then the methylene protons of the benzyl
side chain should remain magnetically equivalent because of this fast rotation.
Consequently freezing the inversion of the N(5) atom would lead to a splitting
of the singlet due to the methylene group into a doublet of probably unequal
population. The actual observed pattern of a quartet (two doublets of equal popu
lation) by Tauscher et al_. (1973) must be due to the fact that the rotation of
the benzyl group is frozen rendering the methylene protons unequivalent, result
ing in a multiplet of an AA'BB' pattern. The published value for the barrier
42.4 kJ/mol Tauscher et al_., 1973) is also in the range expected for the side
chain rotation (Binsch, 1975). Moreover, we could not detect any splitting or
even a broadening in the C spectra of cpd. 6 in acetone down to -95 C.
The symmetry aspects of the dynamics of nitrogen inversion and ring inver
sion will be discussed briefly because these aspects are important for a tho
rough interpretation of the data (Lambert et al_., 1971). Thus if the ring inver
sion ("butterfly") in reduced flavin were followed immediately by nitrogen in
version of the N(5) and the N(10) atoms the mirror image of the initial confor
mation would be formed, resulting in identical NMR spectra. Consequently the ring
inversion in reduced flavin can only be observed in flavin derivatives in which
the inversion barrier is higher than the ring inversion for one of the nitrogen
atoms at least. In cpds. 1 to 4 for instance the ring inversion could only be
studied because of the N(5) acetyl group. In the absence of the acetyl group the
N(5) inversion follows the ring inversion (due to fast proton tunnelling) yield
ing the mirror image of the initial state in agreement with the experimental
observation that NMR spectra of N(5) unsubstituted flavins show no dependence on
temperature. The ring inversion of free 1,5-dihydroflavin, involved in biological
reactions, can therefore only be studied by introduction of sterical hindrance
at the N(5) atom since the N(10) atom will probably have a low inversion bar-2
rier, owing to its partial sp character. It follows from this that it will be
- 68 -
rather difficult to design experiments where the two coupled dynamic processes
13 1 can be studied independently by C and H NMR. With respect to the study of
the N(5) inversion by Tauscher et al_. (1973) it must be mentioned that these
authors did not consider symmetry arguments in the interpretation of their ex
perimental results. Thus, if effects of N(5) inversion have been observed they
have been caused either by a direct N(5) inversion or by ring inversion followed
by N(10) inversion resulting in the enantiomeric form, which cannot be disting
uished from the initial form by conventional NMR methods. The energetically
favoured pathway will of course be dominant. This means that the N(5) inversion
will not be observable if the other pathway is energetically favoured as is the
case in cpd. 6, investigated by Tauscher et al_. (1973).
The existence of mirror-images of flavin has some implications for studies
of flavin and flavin analogues binding by apoflavoproteins. Since, as shown pre
viously (Moonen et aj_. , 1984), in oxidized flavin the N(10) atom is somewhat
out of the molecular plane, an enantiomer also exists in this state. If the apo-
flavoprotein binds the flavin molecule stereospecifically, as can be expected,
then half the amount of an added stoichiometric concentration of flavin with
respect to the apoflavoprotein will be bound faster by the apoprotein than the
other half, because some time will be required for the conversion into the
other enantiomeric form. However, as the conversion will occur with a high rate
(low activation barrier) this aspect will probably only be of importance in
kinetic studies using techniques of high time resolution.
ACKNOWLEDGEMENTS
We are indebted to Mrs. J.C. Toppenberg-Fang for typing the manuscript,
to Mr. M.M. Bouwmans for the preparation of the figures and to Dr. C.A.H.
Rasmussen for carefully reading the manuscript.
69 -
REFERENCES
Anet, F.A.L. and Anet, R. (1975) in: Dynamic Nuclear Magnetic Resonance Spectroscopy (Jackman, L.M. and Cotton, F.A., Eds.) Academic Press, New York, pp. 543-619.
Bernard, M. and St. Jacques, M. (1973) Tetrahedron 29, 2539-2544. Binsen, G. (1975) in: Dynamic Nuclear Magnetic Resonance Spectroscopy (Jackman,
L.M. and Cotton, F.A., Eds.) Academic Press, New York, pp. 45-81. Bushweller, C.H., Anderson, W.G., O'Neill, J.W. and Bilofsky, H.S. (1973) Tetra
hedron Letters, 29, 717-720. Dixon, D.A., Lindler, D.L., Branchard, B. and Lipscomb, W.N. (1979) Biochemistry
18, 5770-5775. Dudley, K.H. and Hemmerich, P. (1967) J.Org.Chem. 32, 3049-3054. Gerig, J.F. (1968) J Am.Chem.Soc. 90, 1065-1066. Grande, H.J., Van Schagen, C G . , Jarbandhan, T., and Müller, F. (1977)
Helv.Chim.Acta 60, 348-366. Jensen, F.D. and Beck, B.H. (1968) J.Am.Chem.Soc. 90, 1066-1067. Kierkegaard, P., Norrestam, R., Werner, P.-E., Csöregh, I., van Glehn, M.,
Karlsson, R., Leijonmark, M., Rönnquist, 0., Stens!and, B., Till berg, 0. and Torbjörnsson, L. (1971) in: Flavins and Flavoproteins (Kamin, H., Ed.), University Park Press, Baltimore, pp. 1-22.
Lambert, J.B. and Oliver, O.L. (1968) Tetrahedron Letters 24, 6187-6190. Lambert, J .B . , O l iver , O.L. and Packard, B.S. (1971) JTAm.uh~em.Soc. 93, 933-937. Lehn, J.M. and Wagner, J. (1970) Tetrahedron Letters 26, 4227-4240. Moonen, C.T.W., Vervoort, J. and Müller, F. (1984) Biochemistry, submitted. Müller, F. and Massey, V. (1969) J.Biol.Chem. 244, 4007-4016. Palmer, M.H. and Platenkamp, R.J. (1979) in: Catalysis in Chemistry and Biochemistry
Reider, Amsterdam, pp. 147-169. Shanan-Atidi, H. and Bar-Eli, K.H. (1970) J.Phys.Chem. 74, 961-963. Simondsen, R.P. and Toi lin, G. (1980) Mol .Cell ,Biochem.~3"3, 13-24. Tauscher, L., Ghisla, S. and Hemmerich, P. (1973) Helv.CïïTm.Acta 56, 630-649. Van Schagen, C G . and Müller, F. (1980) Helv.Chim.Acta 63, 2187~"*~~~].. Van Schagen, C G . and Müller, R. (1981) Eur.J.BioclîëmT T20, 33-39. Werner, P.-E. and Rönnquist, 0 (1970) Acta Chem.Scand. ~~~T 997-1009.
- 70 -
Chapter 7
On The Intermolecular Electron Transfer Between Different Redox States of Flavodoxin
From Megasphaera Elsdenii: A 500 MHz *H NMR Study
Chrit T.W. Moonen and Franz Müller
- 71 -
SUMMARY
The electron transfer reactions between molecules of flavodoxin from
M.elsdenii in different redox states have been investigated by proton nu
clear magnetic resonance techniques at 500 MHz. The electron transfer between
molecules in the oxidized and semiquinone state is shown to be at least 350
times slower than that between molecules in the semiquinone and hydroquinone
state. The latter reaction was studied at different ionic strength and tem
perature. The rate of electron transfer increases with increasing ionic strength,
as expected for a reaction between molecules of identical charges. The electron
transfer reaction is only slightly dependent on temperature suggesting an outer
sphere reaction mechanism. The results indicate that the activation energy
for the electron transfer reaction between the semiquinone and hydroquinone
state is negligible in contrast to that between the oxidized and semiquinone
state. It is suggested that this feature renders M.elsdenii flavodoxin to
an exclusive one electron donor/acceptor in the cell, thereby shuttling be
tween the semiquinone and the hydroquinone state. Mechanistic implications
of the findings are briefly discussed.
- 72 -
INTRODUCTION
Flavodoxins are a class of small flavoproteins (M = 15 000 - 23 000)
which contain riboflavin 5'-phosphate as prosthetic group and function as
electron carriers in biological reactions [1]. Upon reduction these flavoproteins
form a stable semiquinone under anaerobic conditions. The flavosemiquinone
can be reduced further to the two-electron reduced (1,5-dihydroflavin or hydro-
quinone) state (Fig. 1). In in vivo reactions flavodoxin from Megasphaera elsdenii
H„C
H3cy^yNyNY° H 3 C ^ ^ |
Fig. 1. The structures of the prosthetic group in the three redox states.
shuttles between the semiquinone and the hydroquinone state. The redox potential
of this transition is strongly altered in flavodoxins as compared with that of the
free prosthetic group. In M.elsdenii flavodoxin the transition between the quinone
(oxidized) and the semiquinone state (E~) is characterized by a redox potential
of -115mV and that between the semiquinone and the hydroquinone state (E^) by a
redox potential of -375 mV at pH 7.0 [1].
- 73 -
The three-dimensional structure of the flavodoxin from Clostridium MP is known
in all three redox states [2-4]. The crystallographic data show that FMN is shielded
from solvent water except for the two methyl groups at position 7 and 8 of the
isoalloxazine moiety of FMN. No gross conformational change has been observed
between the different redox states, but the three-dimensional data indicate that
a small conformational change takes place in the transition from the oxidized to
the semiquinone state [3].
The chemical and physical properties of the two flavodoxins from M.elsdenii
and Clostridium MP are similar indicating a close structural relationship between
these flavodoxins [1]. This was recently further supported by a nuclear magnetic
resonance (NMR) study [5] showing that the active centers in the two oxidized
flavodoxins are similar. From NMR studies it was also learned that the phosphate
group and the isoalloxazine moiety of FMN are strongly immobilized [6,7]. It was
proposed that the immobilization of the protein-bound flavin may be related to
the biological function of the flavodoxins [7].
It has been proposed from the fact that part of the benzene subnucleus of FMN
in flavodoxin is exposed to the bulk solvent that the methyl group in position
8 of FMN, or less specifically, the 7,8-dimethylbenzene subnucleus of FMN
plays an important role in the electron transfer reaction [8]. Apparent sup
port for this proposal was deduced from kinetic studies using free and protein-
bound modified reduced flavins and cytochrome c as an oxidant. However, the
kinetics of the reduction of cytochrome c by the semiquinone of riboflavin and
7,8-dichlororiboflavin do not support the proposal, since the rate constants for
both reactions are identical [9]. This study, on the other hand, nicely demon
strated that charges play an important role in biological electron transfer re
actions.
In this paper we report on the rate of electron transfer between flavodoxin
molecules of different redox states in equilibrium situations. Although these
reactions are not of physiological importance, we have a relatively simple model
system to investigate the rate of the electron transfer. A major advantage of
- 74 -
the use of the nuclear magnetic resonance techniques in this system is the pos
sibility to follow the flux of electron transfer during equilibrium. The results,
allow us to draw some important conclusions with respect to the mechanism of
electron transfer catalyzed by flavodoxins.
MATERIALS AND METHODS
Flavodoxin from Megasphaera elsdenii was isolated and purified according to
published procedures [10]. Reduction of solutions of flavodoxin was achieved by
the addition of a dithionite solution [11] or by light irradiation for 30 minutes
of flavodoxin solution containing 1.6 mM EDTA and 8 yM deazaflavin [12]. Anaerobic
solutions were prepared by carefully flushing the flavodoxin solution in the
NMR tube with argon for 30 minutes. Stepwise reoxidation was done by injection
of small volumes of air into the NMR tube followed by gentle mixing. During
reduction and reoxidation the intermediate semi qui none state is quantitatively
formed before full reduction or oxidation is accomplished [1].
Samples of flavodoxin solutions were prepared by exhaustive dialysis against
Tris-HCl buffer, pH 8.3. Then the samples were lyophilized and redissolved in
H2O. This procedure was repeated twice and was sufficient to exchange the easily
exchangeable protons by deuterons. The desired concentration of flavodoxin
and buffer were obtained by the addition of the appropriate volume of H~0 to
the final, lyophilized preparation. The ionic strength of the solutions was varied
between 20 to 1000 mM by addition of KCl. The pH values given are those measured
at the beginning of each experiment. It should be noted that the pH values of
solutions reduced by excess dithionite and subsequently reoxidized by addition
of small amounts of oxygen are somewhat lower than the starting pH value due to
the reaction between 0« and the excess dithionite.
H NMR spectra were recorded on a Bruker WM 500 spectrometer operating at
500 MHz. A spectral width of 6000 Hz and a repetition time of 1 s were used.
8K data points were used unless otherwise stated. Quadrature detection and
quadrature phase cycling were used. The accuracy of the temperatures stated
- 75 -
is ± 1 °C. Chemical shifts are reported with respect to trimethylsilylpropionate
(TSP). Wilmad 5 mm NMR tubes were used.
Spin-spin relaxation times were calculated from line width measurements.
Line shape analysis under exchange conditions was performed according to McLaughlin
and Leigh [13] in the extreme conditions of very fast or very slow exchange, and
according to McConnel [14] in intermediate situations (cf. Results). In the
latter case the analytical solution for the two site exchange problem was used
[14,15].
RESULTS
Flavodoxin from M.elsdenii resembles that from Clostridium MP in many
chemical and physical properties [1]. Crystallographic studies of the latter pro
tein have shown that the conformation of the protein is the same in the hydroquinone
and the semiquinone states. With regard to the FMN binding site, however, there are
some minor conformational differences between the quinone state, on the one hand,
and the semiquinone and the hydroquinone states on the other [2-4]. James et al.
[16] proposed, using the H NMR technique, that the overall conformation of M.elsdenii
flavodoxin is almost independent of the redox states (Fig. 1). Van Schagen and
Müller [17] confirmed this conclusion by a H NMR study at higher field strength.
31 These conclusions are further supported by a P NMR study on M.elsdenii flavodoxin
[6].
The high field part of the 500 MHz H NMR spectra of M.elsdenii flavodoxin
in the three redox states is shown in Fig. 2. The resolution of the spectra
is enhanced by a Lorentzian to Gaussian transformation. In spite of the relatively
high molecular weight of the protein (M = 15 000) the resolution is remarkable
and even coupling patterns are clearly seen. This region of the H NMR spectrum
of M.elsdenii flavodoxin consists of methyl resonances strongly shifted from the
random coil positions by ring current effects. These methyl resonances can
therefore be regarded as intrinsic conformational probes of the protein. The
resonance position of most of the lines is not affected by reduction or reoxidation,
except for the resonances I, II and III (Fig. 2). In addition, the line width of
- 76
JSL
-AJ^KJV
A
B
C
D
E
F
G
H
L «—.Afl—AJ^W^__A___
-h-H
-^M~J V * » W ^ ^~«*-'W^«11«»»_»„
I I 1 I I 1 I I I
O -0.4 PPM(Ö)
-0.8
100% HYDROaUINONE
>97% HYDROaUINONE
85% HYDROaUINONE
70% HYDROaUINONE
55% HYDROaUINONE
40% HYDROaUINONE
20% HYDROaUINONE
100% SEMiaUINONE
15% OXIDIZED
45% OXIDIZED
70% OXIDIZED
100% OXIDIZED
Fig. 2. l. The high field part of 500 MHz H NMR spectra of M.elsdenii flavodoxin as dependent on the state of reduction, as indicated. The concentration of the protein was 3 mM. Spectra were taken at 33°C. The protein was dissolved in 20 mM Tris-HCl buffer, pH 8.3. The arrow indicates the resonance position of the internal standard. The roman numbers indicate the resonance lines usee to calculate the rate of electron transfer. For further details, see Materials and Methods, and Results.
- 77 -
the resonances I and III is strongly dependent on the concentration of the semiquinone
present at a certain degree of reduction or reoxidation of the flavodoxin. The
line width of peak II is similarly but less strongly affected by the presence of
semiquinone. The three peaks have already been assigned [17] and represent the
B-methyl group of Ala-56 (peak I) and the two 6-methyl groups of Leu-62 (peak II
and III). Peak I appears at -0.04 ppm in the oxidized and at 0.11 ppm in the
hydroquinone state. Peak III resonates at -0.74 ppm in the quinone and at -0.65
ppm in the two electron reduced state, and peak II at -0.23 ppm in the oxidized
and reduced state.
Flavodoxins form a relatively stable radical under anaerobic conditions [1].
In the semiquinone state the free electron is delocalized in the isoalloxazine
part of the prosthetic group, but most of the spin density is confined to the
central ring (pyrazine subnucleus) of the flavin molecule [18]. The effect of this
paramagnetic center on the resonances in the neighbourhood of the prosthetic group
is seen in spectrum H of Fig. 2. The peaks I and III are broadened almost beyond
detection, whereas the other peaks are relatively unaffected as compared with those
in the diamagnetic redox states (Fig. 2, spectra A and M ) .
On going from the hydroquinone to the semiquinone state peak I and III broaden,
even in the presence of a relatively small concentration of semiquinone (Fig. 2,
spectrum B). As can be deduced from Fig. 2, (spectra A to H) the broadening is
clearly dependent on the relative concentration of hydroquinone and semiquinone.
In contrast, on passing from the oxidized to the semiquinone state (Fig. 2, spectra
M to H) no broadening of the peaks I and III is observed. Only a decrease in peak
height can be detected, the linewidth remains unaltered. The following straight
forward conclusions can be drawn from these observations: 1) when the flavodoxin
molecules are partly in the hydroquinone and partly in the semiquinone state,
the redox state of each flavodoxin molecule changes rapidly with respect to the NMR
time scale (Fig.2, spectra A to G ) ; ii) when the flavodoxin molecules are partly
in the semiquinone and partly in the quinone state, such a rapid change of the
redox state of each flavodoxin molecule does not occur. Similar conclusions can
be drawn from subspectra of all parts of the H NMR spectrum of M.elsdenii flavo-
78 -
7.5 7.0 ».5 6.0
PPM (6)
Fig. 3 The aromatic region of 500 MHz H NMR spectra of M.elsdenii flavodoxin as dependent on the state of reduction. For further details, see Fig. 1. Resonance lines indicated with A are broadened in the semiquinone state. A: 100% hydroquinone, b: 90% hydroquinone, C: 100% semiquinone, D: 70% oxidized, E: 100% oxidized.
- 79 -
doxin in the three redox states. This is illustrated in Fig. 3 for the aromatic
part of the H NMR spectrum. It has been shown by H NMR techniques [5,17] that
several aromatic amino acid residues are in the neighbourhood of the prosthetic
group. The resonances of these amino acid residues behave in the same way as the
methyl resonances at high field. Since the peaks at high field are much better re
solved and separated from other proton resonances the calculation of the rate con
stants for the electron exchange reactions has been done with the experimental
data obtained at high field.
The marked differences between the rate of electron exchange between the dif-
31 ferent redox states has been noticed previously in a P NMR study [6]. Owing to
31
the line width in the P NMR spectra of M.elsdenii flavodoxin and the limited sen
sitivity of the spectrometer used, only approximate rates of the electron exchange
reactions could be determined. On the other hand James et aj. [16] arrived at the
conclusion that the rates of electron exchange are similar for both redox shuttles.
This conclusion cannot be correct, as already indicated above. The discrepancy must
be ascribed to the fact that our H NMR spectra of M.elsdenii flavodoxin are far
better resolved than those of James et al_. [16] who had to base their calculations
on resonance lines composed of several methyl groups.
The spectra were recorded with flavodoxin solutions reduced by various
concentrations of dithionite. In principle one has to consider the possibility that
dithionite or its dissociation product SO«' [11] could function as a mediator in
the electron exchange reactions. To check this possibility the flavodoxin was
reduced by light irradiation in the presence of EDTA and a catalytic amount of
deazaflavin [12]. It was found that the electron exchange rates are independent
on the method of reduction. We therefore can conclude that the measured rate of
electron exchange is due to the direct electron transfer between flavodoxin molecules
in the different redox states. Thus, the line broadening observed in Figs. 2 and 3
- 80 -
represents the following reactions:
Fld^D + Fldhq(2) t n ^ + Fldsq(2) (la) K l
k2
Fid ^ ) + Fid (2) Î Fid ^ + Fid W ( lb) sq q . q sq y '
-2
where Fid denotes M.elsdenii flavodoxin and the subscripts q, sq and hq are the
quinone, semiquinone and hydroquinone states, respectively. Second order rate con
stants are denoted as k. The products of these reactions are the same as the reac-
tants, i.e. the reactions are symmetric and therefore k, = k_,, and k« = k_„.
With the NMR technique it is relatively easy to determine the flux of electrons
during the equilibrium state because one of the reactants is paramagnetic and this
paramagnetism is lost after the electron transfer.
According to the reactions la and lb we are dealing with a simple two-site
exchange system (hydroquinone and semiquinone or semiquinone and quinone).
For this system the total number of electron transfers (in Ms ) can be expressed
as:
nt = kex [A][B] (2)
where [A] and [B] are the concentrations of flavodoxin in the redox states A and B,
respectively, and k is equal to k, or k„ (cf. Eqn. la and lb). The site lifeti
T„ in the redox state A, for instance, is now expressed as the number of mol
cules in the redox state A divided by the number of electron transfers per second
The site lifetimes in the different redox states are related to the lifetime
T [13]:
I - i- + 1 - and <4a> TA TB
x = x A X B = xBXA (4b)
me
e-
- 81 -
where X. and Xß are the mole fractions of flavodoxin, in the redox states A and B.
The combination of Eqns. (2), (3) and (4) leads to the expression:
kex = T _ 1 ( [ A ] + [ B ] ) _ 1 = ^ [ F l d ] " 1 (5)
where [Fid] is the total concentration of flavodoxin in solution. The lifetime T
was determined from the broadening of the resonances I, II and III (Fig. 2). This was
achieved by a fitting procedure of a theoretical lineshape to the experimental line-
shape, using the analytical solution for the two-site exchange problem [14]. In the
limits of slow or fast exchange the equations of McLaughlin and Leigh [13] were used.
The spin-spin (T2) relaxation and the chemical shift of the analyzed atom or
group in the two redox states must be known to simulate the lineshape [14]. T„ was
calculated from the line width. The chemical shift was determined directly from the
H NMR spectra. The parameters for the resonances I, II and III are summarized in
Table 1. The spin-spin relaxation and the chemical shift of resonance I in the
semiquinone state were determined indirectly from spectra A, B and C (Fig. 2) by
extrapolation to 100% semiquinone, using the equations of McLaughlin and Leigh [13]
for the limit of fast exchange.
Table 1. Chemical shift and T_ values of the resonances I, II and III (cf. Fig. 2) of H.elsdenii
flavodoxin in the three redox states
Resonance Redox state
I Ala-56, ß-CH,
II Leu-62, Ô-CH,
III Leu-62, S-CH,
hydroquinone «(ppm) T^Tms)
0.11 26.5±0.2
-0.23 26.5±0.2
-0.65 26.5±0.2
semiquinone <5(PP"i) TJTms)
0.05 0.6±0.2
-0.21 10.6±0.4
-0.70 4.0±0.8
«(ppm) T2(ms) qui none
«(ppm)
0.04 26.5±0.2
-0.23 26.5±0.2
-0.74 26.5±0.2
Under the condition of Fig. 2 the electron transfer between the semi
quinone and the hydroquinone states occurs in the limit of fast exchange and conse
quently only the lower limit of the second-order rate constant can be obtained.
According to Eqn. 5 the electron exchange rate varies proportionally with the
- 82 -
flavodoxin concentration. Experiments at lower flavodoxin concentrations were
therefore also conducted in order to arrive at the intermediate exchange region
for which the fitting procedure allows an accurate determination of the lifetime
and the rate constant. The fitting procedure was done for the resonances I and III
for all experiments, yielding similar results. In situations where a relatively fast
electron exchange (i.e. T < 6 ms) occurs, results were calculated from resonance
I only, because its very small T~ value (Table 1) in the semiquinone state allows
the determination of the lifetime. The slower T 2 relaxation of the group due
to peak III in the semiquinone state (Table 1) does not allow an accurate cal
culation for such short lifetimes. The exchange rates, determined under
various conditions, are summarized in Table 2.
Table 2. Rate of electron exchange between the hydroquinone and semiquinone state
of M.elsdenii flavodoxin under various conditions. Flavodoxin was dissolved
in 20 mM Tris-HCl (pH 8.3). Lifetime T was determined from a comparison
between experimental and computer simulated spectra (cf. Material and
Methods). Rate constants k were determined using Eqn. 5. Accuracy of
the rate constants is ± 20%.
Flavodoxin cone.
(mM)
0.45
0.45
0.45
0.2
0.25
0.25
0.25
Ionic strength3^
(mM)
~5a>
~5a>
200
1000
200
200
200
Temperature
°C
33
33
33
33
8
20
33
Lifetime
T (ms)
17.5
8.5
4.0
<2.5
12
9.5
6.5
Electron exchange
M'V1 x 10b
1.3
2.6
5.6
>20.0
3.3
4.2
6.2
'Ionic strength of the solutions is somewhat higher than indicated. This is partly
due to the addition of dithionite (final concentration ± 1 mM) and partly due to
the lyophilization procedure.
- 83 -
It should be noted that neither a temperature change from 10°C to 33 C nor
a change in ionic strength up to 1 M KCl influences the structure of this flavo-
doxin, as we could not detect any change in the H NMR spectra under these condi
tions. The rate of electron transfer between the semiquinone and the hydroquinone
state is only slightly dependent on the temperature (Table 2.). In contrast the
reaction is strongly dependent on the ionic strength of the solution. The exchange
rate increases with increasing concentration of ionic strength.
3-1 -1 Only an upper limit of the rate constant (<5.7xl0 M s ) could be determined.
for the electron transfer between the quinone and the semiquinone state. Since
this upper limit was calculated under optimal experimental conditions (i.e. 8 mM
flavodoxin, IM KCl at 33°C) we had no choice to perform meaningful experiments
under different experimental conditions.
The distance to the paramagnetic center can be calculated in the semiquinone
state from the T~ values of the groups due to the resonances I, II and III according
to the Solomon-Bloembergen equation [19,20]:
1 1 vn2fi2 3 T R 1 3 T R
T ' i V - (4TR + 777^—2 + - 7 ^ - ^ <6> 2 20 ru 1+ (UH T R r 1 + ( " V R )
where Y H is the gyromagnetic ratio of hydrogen, g the electron g-value, ß the
Bohr magneton, r the distance between the hydrogen atom and the free electron,
io<- and DM respectively the electron and hydrogen Larmor precession frequencies,
T R the correlation time, and S the spin quantum number. The correlation time
consists of contributions from the rotational correlation time T and the c
electron relaxation time T :
k- = L- + L (7)
TR Tc Ts
The rotational correlation time of the protein is about 5 ns [7]. T is about
100 ns [21] and can thus be neglected in the calculations. Only the dipolar part
of the Solomon-Bloembergen equation is considered in Eqn. 6 because no electron
- 84 -
density of the free electron is delocalized onto the groups under consideration.
The following values are calculated for the distance between the paramagnetic center
and Ala-56 (peak I ) , 0.73 rati; Leu-62 (peak II), 1.02 nm; Leu-62 (peak III), 1.28 ntn.
The following distances, with respect to the N(5) atom of the isoalloxazine ring,
were determined from crystallographic data of the related flavodoxin from Clostridiu
MP [2]: 0.49, 1.12 and 1.28 nm, respectively. Our calculated values agree better
with the crystallographic data as the distance increases. The reason for this
effect probably lies in the fast rotation of the methyl groups, for which Eqn. 6
does not account. The effect of fast rotating methyl groups has been clearly demon
strated by Kalk and Berendsen [22]; the closer the methyl groups to the paramagnetic
center, the more the angle of the rotating vector increases. Consequently, the
influence of the fast rotation of the methyl groups to the relaxation rates in
the semiquinone state increases when the distance to the paramagnetic center
decreases. Furthermore, it cannot be excluded that the conformation of the active
center of M.elsdenii flavodoxin differs somewhat from that of Clostridium MP flavo
doxin.
DISCUSSION
The maximum electron exchange rate for the electron transfer between the
semiquinone and the hydroquinone state is observed at a high salt concentration (Tab
2). Under the same conditions (I=1M, 33°C) the upper limit for the electron
transfer between the qui none and the semiquinone state was found to be at least
3-1 -1 350 times slower (k < 5.7 x 10 M s ) than the electron exchange between the
semiquinone and the hydroquinone state. As we have measured the flux of electrons
during equilibrium and identical conditions, we conclude that this difference must
be due to a difference in activation energy, i.e. the activation energy for the
electron transfer between the semiquinone and the hydroquinone state is markedly
lower than that for the electron transfer between the semiquinone and the quinone
state. A lower limit of the difference of the activation energy for the two elec-
- 85 -
tron exchange reactions can be calculated from the difference of the two rate
constants. This yields AE > 15 kJ/mol. The activation barrier between the semi-
quinone and the hydroquinone state can be calculated from the temperature dependence
of the rate constant (Table 2). An increase of the reaction rate by a factor of
1.9 ± 0.4 is observed on going from 8°C to 33°C. From these results an activation
energy of 18.5 kJ/mol can be calculated. However, an increase in temperature leads
to a decrease in the viscosity of water and consequently to an increase in the
diffusion constant by a factor of 1.85. This strongly indicates that the acti
vation energy for the electron transfer between the semiquinone and the hydroquinone
state is rather small, if not negligible. This means that the reaction is dif
fusion controlled and strongly suggests that we are dealing with an outer sphere
reaction mechanism [23]. From the apparent absence of an activation energy it
must be concluded that the rate of electron transfer is determined by the number
of times that the active sites are in contact with each other.
Flavodoxin from M.elsdenii has a net charge of -17 at pH 7 as calculated
from the primary structure [24,25]. It is known from three-dimensional data of
the related flavodoxin from Clostridium MP that most of the net charge is con
centrated near the FMN binding site [2-4]. In addition the C-terminal of a helix
dipole moment points towards the prosthetic group. These facts suggest that a
strong repulsion between flavodoxin molecules exists. This suggestion is supported
by data of Moonen and Muller [6] showing that no apparent association occurs between
flavodoxin molecules, even at a concentration of 3 mM. The strong ionic strength
dependence of the exchange rate is in accordance with the latter results (Table 2)
and with results of Simondsen et aj_. [26], who showed that the electron transfer
rate between the positively charged cytochrome c and flavodoxin decreased with
increasing ionic strength. At IM KCl the repulsive forces of the charges are ap
proximately totally diminished and under this condition only a lower limit of the
exchange rate between flavodoxin molecules could be determined.
The absence of an activation barrier for the electron transfer reaction be-
- 86 -
tween the semiquinone and the hydroquinone state allows us to compare directly
the experimentally determined exchange rate (k ) with the calculated collision
frequency (k ) in order to determine the efficiency of the reaction. Mathematically
this can be expressed with the following equation:
kex = 0.5 k c x f(Dr o t, 0 e f f , r, p) (8)
where f is a function of the effective surface (0 f f ) , the distance needed
for an effective transfer and the geometrical constraints for the reaction
(factor p). Van Leeuwen [27] has pointed out that rotation of the reacting mole
cules during the encounter time increases the efficiency of the reaction (D . in
Eqn. 8). This term is probably of minor importance in our case since the disso
ciation rate of the complex must be rather high. A lower limit for the dissociation
rate, respectively an upper limit for the lifetime of the complex can be estimated
neglecting the effect of charges. Assuming that equilibrium conditions exist
when a certain volume contains the theoretically maximal amount of flavodoxin
which is calculated to be 67 mM. Under these conditions the flavodoxin molecules
are forced to interact with each other. It is then calculated [27] that the life
time of the complex is < 1.7 ns.This value must be considerably smaller owing to
the existing repulsion forces (see above). Since the rotational correlation time
T of the protein is 5 ns [6] it can be concluded that D . (Eqn. 8) can indeed
be neglected. This conclusion is further supported by the facts that no association
between flavodoxin molecules was observed [6], that the dipole moment of the
molecules is situated such that repulsive forces develop on approach of the active
sites (Van Leeuwen, personal communication) and that the negative charges in
the flavodoxin molecule are concentrated around the active site. These latter
repulsive forces are not diminished at a high salt concentration since by
a close approach of the flavodoxin molecules counterions are forced out of the
intersecting area. The factor 0.5 in Eqn. 8 takes into consideration that even
under ideal conditions the electron might either be transferred or stay on the
- 87 -
same molecule. If we assume that there are no geometrical constraints for the
reaction and that a very close contact is needed (i.e. r ~ o) we can de
termine the effective area of the reaction using Eqn. 9:
2
(9) kex = ° - 5 k c °eff
°tot/
on
where 0 .. and 0. . are the effective and the total surface of the protein,
respectively. The power 2 originates from the fact that both active centers
have to meet each other for an effective electron transfer. The collision fre
quency k can be calculated from Eqn. 10 [27] under the condition that charges
can be neglected. The latter condition is sufficiently fulfilled at 1 M KCl.
kc = 4 7 r N A R D x l ° 3 (10)
In Eqn. 10 N. is the number of Avogadro, R the sum of the radii of the two
colliding flavodoxin molecules (r, = r~ = 1.8 nm [2]), D the sum of the diffusi ?n -7 ? -i ^
constants of the two flavodoxins (D = 11.4 x 10 c m s [10]) and 10 a
scaling factor. A collision frequency of 8.7 x 10 M~ s" is calculated at 33°C
fill from Eqn. 10. Using this value and the lower limit of k (2.0 x 10 M s " ) and a
o -. o
total surface of 4.07 x 10 8 we arrive, under the condition of Eqn. 9 at an
effective surface of 87 8 , i.e. 2.1% of the total surface. The calculated
effective surface would increase if geometrical constraints for the reaction
existed. If electron transfer, on the other hand, would also be possible at larger
distances, the effective surface would decrease (cf. Eqn.8). The calculated value
should therefore be regarded with caution. Nevertheless, we can conclude that
a relative large effective surface is needed or the electron transfer is possible
without a very close approach of the molecules. It should be noted that both pos
sibilities argue against the view that the fairly exposed methyl groups at C(7)
and C(8) of the prosthetic group have to be in direct contact with the electron
acceptor or donor. This notion is supported by electron transfer experiments using
7-methyl-10-ribityl-isoalloxazine-5'-phosphate as a prosthetic group. M.elsdenii
flavodoxin reconstituted with this flavin yielded identical electron transfer
rates (data not shown) to those obtained with the native protein. This result
seems to be in contradiction with the conclusion of Simondsen and Toi lin [9]
that the dimethyl benzene subnucleus is directly involved in the electron transfer
reaction. This conclusion was mainly based on kinetic experiments using 8-chloro
substituted flavins. However, such substitutions might introduce an activation
barrier in the redox transitions. Another possibility is that the chloro substituted
flavins have less orbital overlap with the acceptor or donor. Moreover, the
vibrational modes of flavin are strongly altered upon this chlorine substitution
[28]. But sufficient orbital overlap is a condition in an outer sphere mechanism
following Franck-Condon principles [23].
The rate constants of the electron transfer reactions for M.elsdenii flavo
doxin, determined under equilibrium conditions, are in excellent agreement with
kinetic studies by Mayhew and Massey [29] , using stopped flow techniques. Owing
to the high reactivity of the semiquinone of M.elsdenii flavodoxin towards di-
thionite the rate of the reduction of the semiquinone form could not be deter
mined, but it was estimated that the semiquinone is about 450 times more reactive
than oxidized flavodoxin. The rate constant for the comproportionation reaction,
i.e. the reaction between flavodoxin molecules in the oxidized and reduced state
to form semiquinone, was determined to be 2.3 x 10 M~ s" (1=0.16 M, 25°C).
Considering the different reaction conditions this value is in excellent agree
ment with our maximum value of 5.7 x 10 M~ s" (1=1.0 M, 33°C) for the elec
tron exchange reaction between molecules in the oxidized and the semiquinone
state. This means that the comproportionation reaction is solely governed by the
rate of the one-electron reduction of oxidized flavodoxin, i.e. the activation
energy of the transition from the oxidized to the semiquinone state.
The conformation and the conformational stability of flavodoxin in the three
redox states [2-7] can also be rationalized by an outer sphere reaction mechanism.
In an outer sphere reaction mechanism the electron is transferred from one
- 89 -
molecule to the other, while no nuclear rearrangement occurs according to
the Franck-Condon principles. Thus the first step will be an electron transfer
from the HOMO of the reduced state to the LUMO of the semi qui none state. The
next (much slower) step is the electronic relaxation, accompanied by nuclear
rearrangement to the newly formed hydroquinone and semi qui none state. It is
well known that during the first step (the electron transfer) the polarity of
both reactants are considerably changed. As a consequence polar residues and
especially solvent water have to rearrange and, therefore, a high activation
energy can be introduced. This is the case for the transition quinone-semiquinone,
where a conformational change was observed in the FMN binding site, using crystallo-
graphic methods [3]. In the semiquinone and hydroquinone states however, the con
formation is identical. Moreover, the very tight package of the active center of
flavodoxin [5-7] shields the flavin ring from solvent water. Thus nuclear rear
rangement is considerably hampered. This rationalizes the tight and identical
structures of the semiquinone and the hydroquinone states, whereas the introduction
of a considerable nuclear rearrangement during the quinone/semiquinone transition
inhibits the involvement of the oxidized state in the electron transfer. This
mechanism probably renders flavodoxin from M.elsdenii to an exclusive one-electron
transferring protein. Furthermore, our results place some doubt on the suggestion
that specific complex formation is needed between donor and acceptor. We have
given evidence in this paper, that considerable transfer rates can be observed
without any specific complex formation. Specific complex formation might of course
enhance the transfer rate but it seems not to be a prerequisite for the electron
transfer reaction.
It should be noted that electron carriers as flavodoxins and ferredoxins
are often interchangeable, although structures and chromophores are quite
different. Mayhew and Ludwig [1] explained the interchangeability as an apparent
lack of recognition in the electron transfer reactions. This explanation fits
well with our suggestion, that tight and specific complex formation is not a
- 90 -
necessary condition for the electron transfer. We suggest, that this is es-
specially valid when a negligible activation barrier exists for the involved
redox transitions. If a high barrier exists, than specific complex formation can
be necessary in order to overcome the barrier.
From our data it is likely that the difference in activation energy is
the control mechanism in the living cell for the preference for the shuttle be
tween the hydroquinone and the semiquinone state above that between the semiquinone
and the quinone state. In this respect it is interesting to note that M.elsdenii
flavodoxin is an inducible protein which replaces ferredoxin in the absence of
iron [1]. Ferredoxin transfers one electron with a midpoint potential of -375 mV
[30]. The redox potential of the semiquinone/hydroquinone couple of flavodoxin
is -375 mV whereas the midpoint redox potential of the quinone/semiquinone couple
is -115 mV [1,30]. From the comparison of these redox potentials and the inter-
changeability of the two proteins it has already been deduced [1] that the function
of M.elsdenii flavodoxin is likely to shuttle between the hydroquinone and the
semiquinone state. In this study we have found strong support for this suggestion.
It is questionable, however, if this feature is common to all flavodoxins as
some flavodoxins are constitutive. Thus no clear interchangeability with a one-
electron transferring protein exists for the other class of flavodoxins. Thus
it cannot be excluded that these flavodoxins can also act as two electron donors
or acceptors. This aspect is currently under study in our laboratory.
- 91 -
ACKNOWLEDGEMENTS
We are indebted to Miss C M . Verstege for typing the manuscript, to
Mr. W.A.M. van den Berg for the isolation of the flavodoxin, to Mr. M.M. Bouwmans
for the preparation of the figures, to Dr. C.A.H. Rasmussen for carefully reading
the manuscript and to Dr. J.W. van Leeuwen for valuable discussions.
This study has been carried out under auspices of the Netherlands Foundation
for Chemical Research (SON) with financial aid from the Netherlands Organization
for the Advancement of Pure Research (Z.W.O.).
REFERENCES
1. Mayhew, S.G. and Ludwig, M.L. (1975) The Enzymes, 3rd Ed. 12, pp. 57-118. 2. Burnett, R.M., Darling, G.D., Kendall, D.S., LeQuesne, M.E., Mayhew, S.G.,
Smith, W.W. and Ludwig, M.L. (1974) J.Biol.Chem. 249, 4383-4392. 3. Smith, W.W., Burnett, R.M., Darling, G.D. and Ludwig, M.L. (1977)
J.Mol.Biol. 117, 195-225. 4. Ludwig, M.L., Burnett, R.M., Darling, G.D., Jordan, S.R., Kendall, D.S.,
Smith, W.W. (1976) in: Flavins and flavoproteins (Singer, T.P., Ed.) Elsevier, Amsterdam, pp. 393-404.
5. Moonen, C.T.W., Höre, P.J., Müller, F., Kaptein, R. and Mayhew, S.G. (1982) FEBS Lett. 149, 141-146.
6. Moonen, C.T.W. and~Wller, F. (1982) Biochemistry 21, 408-414. 7. Moonen, C.T.W. and Müller, F. (1983) Eur.J Biochem. 133, 463-470. 8. Simondsen, R.P. and Tollin, G. (1983)~B~Töchemistry 27T3008-3016. 9. Ahmad, I.,- Cusanovich, M.A. and Tollin, G. (1981) Proc.Natl .Acad.Sei. USA
78, 6724-6728. 10. Mayhew, S.G. and Massey, V. (1969) J.Biol.Chem. 244, 794-802. 11. Mayhew, S.G. (1978) Eur J.Biochem 85, 535-547. ~~ 12. Massey, V. and Hemmerich, P. (1978J~Biochemistry 17, 9-17. 13. McLaughlin, M.A. and Leigh, J. (1973) J.Magn.Reson. 9, 296-304. 14. McConnell, H.M. (1958) J.Chem Phys. 28, 430-435. 15. Jardetzky, 0. and Roberts, G.C.K. (138~1) in NMR in Molecular Biology,
Academic Press, New York, pp. 115-142. 16. James, T.L., Ludwig, M.L. and Conn, M. (1973) Proc.Natl.Acad.Sei. USA
70, 3292-3295. 17. Van Schagen, C G . and Müller, F. (1981) FEBS Lett. 136, 75-79. 18. Müller, F., Hemmerich, P., Ehrenberg, A., Palmer, G. and Massey, V.
(1970) Eur.J.Biochem 14, 185-196. 19. Solomon, I. (1955) Phys.Rev. 99, 559-565. 20. Bloembergen, N (1957) J.Chëiïï.Pïïys. 27, 572-573. 21. Visser, A.J.W.G., de Wit, J., Müller, F. and Berendsen, H.J.C (1982)
Helv.Chim.Acta 65, 2422-2430. 22. Kalk, A. and Berendsen, H.J.C. (1976) J.Magn.Reson. 24, 343-366. 23. Devault, D. (1980) Quart.Rev.Biophys. 13, 387-564. 24. Tanaka, M., Haniu, M., Yasunobu, K.T.,"Mayhew, S.G. and Massey, V.
(1973) J.Biol.Chem. 248, 4354-4366. 25. Tanaka, M., Haniu, M77~Yasunobu, K.T., Mayhew, S.G. and Massey, V. (1974)
J.Biol.Chem. 249, 4397. 26. Simondsen, R.P., Weber, P.C., Salemme, F.R. and Tollin, G. (1982)
Biochemistry 21, 6366-6375. 27. Van Leeuwen, J.W. (1983) Biochim.Biophys.Acta 743, 408-421. 28. Schopfer, L.M. and Morris, M.D. (1980) Biochemistry 19, 4932-4933. 29. Mayhew, S.G. and Massey, V. (1973) Biochim.Biophys.Acta 315, 181-190. 30. Van Dijk, C , van Leeuwen, J.W., Veeger, C , Schreurs, J.P.G.M. and Baren
drecht, E. (1982) J.Electroanal.Chem. 9, 743-759.
- 92 -
Chapter 8
13 15 A C and N Nuclear Magnetic Resonance Study on The Interaction between
Riboflavin and Riboflavin-Binding Apoprotein
Chrit T.W. Moonen, Willy A.M. van den Berg, Marleen Boerjan and Franz Müller
- 93 -
ABSTRACT
Riboflavin-binding apoprotein from egg yolk and egg white have been 15 13
reconstituted with N and C enriched riboflavin derivatives. These protein preparations were investigated in the oxidized and two-electron re
ib 13 duced state by N and C NMR techniques. The chemical shift values of the
protein-bound flavin are compared to those of free flavin in water or chloro
form. The results are interpreted in terms of interactions between the apo
protein and its prosthetic group and the conformation of bound flavin.
In the oxidized and in the reduced state the chemical shifts and line
broadening effects of the bound riboflavin are extremely similar for the two
different proteins indicating similar binding sites for riboflavin. The sol
vent accessibility to the bound riboflavin is similar in the reduced and the
oxidized complexes, i.e. the N(3), 0(2a) and N(l) atoms are exposed, whereas
the 0(4a) atom is shielded from solvent water.
A hydrogen bond to the N(5) atom exists in the oxidized state. This hy
drogen bond is weaker than the corresponding one between free FMN and water.
At best only a weak hydrogen bond exists to the 0(4a) atom. Binding of oxidized
riboflavin forces the N(10) atom of flavin into the molecular plane. The iso-
alloxazine ring is strongly polarized.
No hydrogen bond to the 0(4a) atom is formed in the reduced neutral state
of the complex. Although the N(l) atom is accessible to solvent water, it pos-3
sesses more sp character than in free FMNHL. The data also suggest a rapid
motion of the N(l) atom in and out of the molecular plane. The reason for these
effects is probably the absence of a hydrogen bond to the 0(4a) atom which
makes -n electron derealization from the N(l) atom unfavourable. The N(10) atom
exhibits about the same configuration as in the oxidized state. In contrast the 3
N(5) atom shows increased sp hybridization.
The pK value of the deprotonation of the N(1)H group is measured as 7.45,
a shift of about 0.8 pH unit to more alkaline pH values as compared to that of
free FMNH0. This pK shift might be caused by the absence of a hydrogen bond to e a 3
the 0(4a) atom. In the anionic reduced state the N(l) atom shows even more sp
character than in the neutral complex, which is explained by the absence of a
- 94 -
15 hydrogen bond to the 0(4a) atom. The N resonance line due to the N(l) atom is
severely broadened, probably due to its motion in and out of the molecular plane.
15 The N(5) atom shows almost the same N chemical shift as in the neutral complex 3
indicating an increased sp character. This is confirmed by the coupling constant
of 78 ± 5 Hz with the covalently bound hydrogen atom. No coupling was detected
in the neutral complex due to rapid hydrogen exchange with solvent water. These
results show that the N(5) atom is shielded from solvent at high pH values. The
N(10) atom exhibits considerable sp character in the anionic complex.
The results are discussed in the light of the possible function of the
protein.
INTRODUCTION
Riboflavin-binding protein (RBP) is found in the yolks and whites of
all avian eggs (Osuga and Feeney, 1968). The function of the protein is still
not fully understood, although it is known, that it mediates the transport
of riboflavin from the maternal system to the developing oocytes, where
riboflavin is essential for embryonic growth and development (Winter et al.,
1967; Ostrowski et aj_., 1968). It has also been suggested that RBP in egg
whites functions as an inhibitor of bacterial growth (Board and Fuller, 1974).
As flavodoxins (Mayhew and Ludwig, 1975) RBP is often used as a model
system for flavin-protein interactions. The easy availability of large quan
tities of these proteins makes them especially suitable for a study of the
flavin-apoflavoprotein interactions. The importance of these interactions in
flavin biochemistry is obvious since it has been suggested that specific in
teractions in flavin-apoflavoprotein complexes are responsible for the "tuning"
of the flavoprotein for its specific catalysis (Müller, 1972; Eweg et al_., 1982).
Several physicochemical studies have been reported on RBP and its inter
action with riboflavin. It has been shown that RBP from egg yolk (RBPY) resem
bles RBP from egg white (RBPW) with respect to the thermodynamics of riboflavin
binding (Matsui et al_., 1982a, 1982b). The two proteins have roughly the same
- 95 -
molecular weight (Ostrowski and Krawczyk, 1963; Blankenhorn, 1978; Froehlich
et al_., 1980), the same amino acid composition and flavin-protein association
constants (Farrell et a]_-, 1969; Ostrowski and Krawczyk, 1963; Steczko and
Ostrowski, 1975), and are immunologically indistinguishable (Farrell et al.,
1970). They differ however with respect to the content of sialic acid (Jacub-
czak et al_., 1968; Kawabata and Kanamori, 1968). In addition Matsui et al.
(1982a and 1982b) have shown that the volume of the cavity for the binding
site of riboflavin differs for the two proteins. This has however no influence
on the kinetics of the association between flavin and apoprotein, as shown by
Nishina (1977).
The use of riboflavin analogues provided new insights into the relative
importance of the different atoms of the isoalloxazine ring for the binding
by RBPY and RBPW (Matsui et a]_., 1982a and 1982b; Choi and McCormick, 1980;
Becvar and Palmer, 1982; Nishikimi and Kyogoku, 1973). These studies resulted
in the conclusion that the dimethyl benzene ring of flavin is buried in the
interior of the protein, i.e. bound differently than in flavodoxins (Mayhew
and Ludwig, 1975). Thermodynamic studies on RBPW support this conclusion
(Matsui et al_., 1982a). In addition it was shown by chemical modification
studies that a tryptophan and a tyrosine residue are involved in the binding
of riboflavin by RBPW (Blankenhorn, 1978) and probably also by RBPY (Steczko
and Ostrowski, 1975). Kumosinski et al_. (1982) suggested for RBPW that an
aromatic-rich cleft opens at pH values 3.7<pH<7.0 and concommitantly releases
the riboflavin.
13 15 The C and N NMR techniques are powerful methods for detecting subtle
differences in hydrogen bonding and conformation of the flavin in both the
oxidized and the reduced state (Moonen et a K , 1984) and for elucidation of dy
namic processes (Moonen and Müller, 1983). These methods have successfully been
13 15 applied, using specifically C and N enriched flavins, to flavodoxins from
Megasphaera elsdenii and Azotobacter vinelandii (Van Schagen and Müller, 1981;
13 Franken et al_., 1983), where specific hydrogen bonding was detected. Some C
NMR experiments on oxidized RBPW were carried out by Yagi et al. (1976). No
- 96 -
theory for the interpretation of the data was developed at that time. We now
report on a detailed study on the complex between RBPY and riboflavin, and RBPW
13 15 and riboflavin using specifically labelled C and N riboflavin derivatives.
Several specific interactions are detected in both the oxidized and the reduced
systems. In addition novel information is revealed concerning the hydrogen ex
change of hydrogen-carrying nitrogen atoms, which show an unexpectedly strong
dependence on pH. Dynamic information about the riboflavin binding site is de
duced from these results.
MATERIALS AND METHODS
Apo-RBPY was isolated and purified according to Murthy et al_. (1979)
with the exception that an Ami con concentrator was used instead of aquacide
treatment. Apo-RBPW was isolated and purified according to Farrell et al.
13 (1969). Specific C enrichment of riboflavin was performed according to
15 Van Schagen and Muller (1981). Specific N enrichment was performed as described by Müller et al_. (1983).
The relative concentrations of riboflavin-binding apoprotein and riboflavin
was 1:1 in all samples. These were prepared as follows. Reconstitution of ribo
flavin-binding apoprotein with riboflavin was carried out with 0.1 mM apoprotein
and an excess of riboflavin in 50 mM potassium phosphate (pH 7.0). Excess of
riboflavin was removed by exhaustive dialysis against the same buffer. In the
final dialysis the buffer concentration was decreased to 1 mM. The protein-ribo-
flavin complex was then concentrated by lyophilization.
Wilmad 10 mm precision NMR tubes were used. The samples contained 1-3 mM
RBP, 100 mM potassium phosphate buffer of varying pH (see Results and Discus
sion). Depending on the kind of experiment, the samples contained 10% or 2
100% H20. The sample volume was 1.6 ml. The pH measurements were carried out
before and after the NMR measurements and are reported without correction for
the deuteron isotope effect. Reduction was conducted by addition of the desired
amount of a dithionite solution to the anaerobic solution of RBP. Anaerobiosis
- 97 -
was achieved by carefully flushing the solutions in the NMR tube with argon
for about 20 min. The NMR tube was sealed with a serum cap.
All measurements were performed on a Bruker CXP 300 spectrometer operating
at 30.4 MHz for N and 75.6 MHz for 1 3C NMR experiments. Broadband decoupling
13 of only 0.5 Watt was applied for C measurements to prevent heating up of the
13 sample. There was no need for a higher decoupling power for C NMR as we were
15 mainly interested in quaternary carbon atoms. No decoupling was used for N
NMR measurements in order to be able to observe the exchange rate of the NH
13 protons. Dioxane (3 ul) was used as an internal reference for C NMR. Chemical shift values are reported relative to TMS (^Hioxan'^TMS = 67-84 PP m )•
15 15
Neat [ N]CH3N02 was used as an external reference for N NMR using a coaxial
cylindrical capillary as recommended by Witanowski et al_. (1981). Chemical
shift values are reported relative to liquid NH, at 25°C as recommended by
Levy and Lichter (1979) (óCH NQ -ôN„ = 381.9 ppm for the magnetic field paral
lel to the sample tube (Witanowski et al_., 1981)). Values are reported as true
shiel dings, i.e. corrected for bulk volume susceptibilities as described by
Witanowski et al_. (1981). The accuracy of the reported values is ±0.1 ppm for 13 15
C and ±0.3 ppm for N chemical shift values, unless otherwise stated. The
temperature of all samples was kept constant at 26±2°C. All spectra were
recorded using 30° pulses and a repetition time of 0.5-1 s.
RESULTS AND DISCUSSION
§îydies_on_the_oxidized_state
The dissociation constant of RBPW and RBPY with riboflavin is about
50 nM (Choi and McCormick, 1980; Zak et al_., 1972). For the samples used in
this study it is calculated that only about 0.1% riboflavin is free in solu
tion. Consequently the reported data are not distorted by free riboflavin.
13 Spectrum A of Figure 1 shows the low field part of the C NMR spectrum of RBPY
complexed with riboflavin. The possible interference with protein resonances for
the determination of the chemical shift values of bound riboflavin was elimi-
- 98 -
C(4a) C(2)
vw^^^
180 160 140
PPM (6)
120 100
Figure 1: 13C NMR spectra of r ibof lav in-b ind ing apoprotein from egg yolk reconsti tuted with [ 2 ,4a- 1 3 C 2 ] r ibo f lav in . Spectrum A: oxidized protefn, in 100 mM potassium phosphate, pH 8.5. Spectrum B and C: two-electron reduced protein in 100 mM potassium phosphate, pH 6.0 and in 100 mM sodium borate, pH 9.0, respectively.
nated by the use of difference spectra (data not shown). The resonances due to
13
natural abundance C of the protein in the spectra can partly be assigned ac
cording to Allerhand (1979) and Gratwohl and Wüthrich (1974). All peptide carbonyl
and carboxyl resonances appear between 170 and 180 ppm. The Arg C ç atoms are
- 99 -
found at 158.2 ppm. Aromatic carbon atoms of Trp, Tyr, Phe and His resonate be
tween 110 and 140 ppm. An interesting region lies between 95 and 105 ppm where
normally no carbon atoms of amino acids resonate. However several natural abund
ance resonances of RBPY are found in this region. As already pointed out in the
introduction, RBPY contains sialic acid. The C-l atoms of some carbohydrate com
ponents of sialic acid appear in the region between 95 and 105 ppm. The resonance
position should in principle allow a rough determination of the different stereo
isomers (Bock and Thogerson, 1982). In addition, several lines due to other car
bohydrate residues were observed in the spectral region between 60 en 80 ppm
(data not shown). It should be mentioned that these regions differ between RBPY
and RBPW, as would be expected from the difference in sialic acid content. How
ever, these spectral regions are not further considered here as this lies beyond
the scope of this study.
13 The flavin C chemical shifts for the complexes of RBPY and RBPW are sum-
15 15
marized in Table I. Spectrum A of Figure 2 shows a N NMR spectrum of N
enriched riboflavin bound to RBPW. Only small natural abundance resonances near
120 ppm are observed which are due to peptide nitrogens (Levy and Lichter, 1979). 15 The N chemical shifts are also summarized in Table I. The structure of the
15 flavin labeled at all four nitrogen atoms with N is given in Figure 3.
In the accompanying paper (Moonen et al_., 1984) we have shown that the
13 15 C and N chemical shifts are very sensitive to hydrogen bonding, conformation
of the flavin and the polarity of the solvent used. The extreme situations of
strong hydrogen bonds and high polarity around the isoalloxazine ring (e.g. FMN
in water) as against no hydrogen bonds and low polarity (e.g. TARF in chloroform),
are also listed in Table I in order to facilitate the analysis.
A rough comparison between the data of riboflavin bound to RBPY and RBPW
shows (Table I) that the chemical shifts for the two proteins hardly differ. The
differences are all within the experimental accuracy of the data. This leads to
the conclusion that the binding site for the oxidized isoalloxazine ring must be
very similar in the two proteins. Consequently we can discuss the interactions
for both complexes together.
- 100 -
Table I. Carbon-13 and nitrogen-15 chemical shifts of oxidized riboflavin bound to riboflavin-binding apoprotein from egg yolk (RBPY) and from egg white (RBPW). Linewidth of '3C resonances is 10±4 Hz and that of 15N resonances 15±5 Hz, unless otherwise stated. For comparison also 13C and 15N chemical shifts of FMN in water and of TARF in C?HC13 are included.
Compound Solvent Chemical shifts in ppm
RBPWa)
RBPY3)
FMNb'c)
TARF
RBPWa'd)
RBPWd'e)
R B P Ya 'd)
MeIMNb'c'
MeTARId)
pH 8.5
pH 8.5
pH 8.0
C2HC13
pH 9.0
pH 6.4
pH 6.2
d)pH 8.0
C2HC13
Carbon Atoms
C(2)
159.7
159.6
159.8
155.2
N
N(l)
191.5
191.5
191.6
190.5
201.1
trogen
C(4)
162.1
162.0
163.7
159.8
Atoms
N(3)
159.2
158.5f)
159.4f)
160.4
160.7
C(4a)
135.9
135.9
136.2
135.5
N(5)
338.2
337.7g)
338.2g)
335.5
346.7
C(10a)
152.3
152.4
152.1
149.1
N(10)
165.4
165.1
165.3
164.6
150.4
a) Solvent is 90% H20, 10% H20.
b) FMN or MeIMN were used for the comparison because of their higher solubility in HpO than riboflavin analogues
c) Independent of pH in the range 5<pH<9
d) See Figure 3 for structure of the flavin
e) Solvent is 100% ^ 0
15» f) The resonance signal of N(3) is broadened at a pH value of about 6, linewidth <\40 Hz. Accuracy of the chemical shift given is ±1 ppm (See text).
g) Linewidth is about 30 Hz.
- 101 -
Contrary to H NMR where ring current effects of neighbouring aromatic amino
acids influence chemical shifts such effects are of relative negligible importance
13 15 in C and N NMR. Therefore, the data in Table I can be analyzed directly in
terms of hydrogen bonding, specific interactions and conformation of the flavin
in the complexes following the semi-empirical rules of the accompanying paper
(Moonen et al_., 1984).
15 In the oxidized state the N(3) atom in riboflavin bound to the two proteins
is the only nitrogen atom carrying a covalently bound hydrogen atom.
A J,,- , of 92 Hz is expected (Franken et al., 1983). No splitting of the 15 ^ ( 3 ) - ^ -
N signal due to the N(3) atom of flavin is observed on use of HJ) as a solvent.
This indicates that the exchange rate of the proton is high (lifetime T <<•>-,
i.e. T « 1 0 ms). This result strongly suggests that the N(3) atom is exposed to
solvent and also explains why RBP can be easily purified using riboflavin coupled
via the N(3) position to a matrix (Blankenhorn et al_., 1975). Choi and McCormick
(1980) arrived at the same conclusion by studying the interaction between RBP
and flavin analogues.
The resonance position of C(2) is almost the same as that of FMN in water.
This suggests that 0(2a) is also exposed to solvent water, besides N(3), again
in accordance with the binding studies of flavin analogues. The resonance posi
tion of C(4) clearly shows that the hydrogen bond to 0(4a) is considerably
weaker than in free FMN. The upfield shift of N(3) is probably a consequence
of the weak hydrogen bond to 0(4a), as by approximation the N(3)H-C(4)-0(4a)
group in oxidized flavin can be regarded as a "peptide bond". It is known that
the nitrogen resonance of peptides is mainly governed by the hydrogen bond to
the carbonyl function (Witanowski et al_., 1981). The resonance position of
N(l) indicates that a strong hydrogen bond is formed to the N(l) atom of bound
riboflavin. This hydrogen bond is comparable to that of FMN in water. Thus,
with regard to hydrogen bonding and solvent exposure, the data indicate that
N(3), 0(2a) and N(l) are exposed to solvent, whereas N(5) exhibits a somewhat
weaker hydrogen bond as compared with free FMN.
- 102
N(5)
l*fWj
NI3) N(10)
l ^ ^ ^ j V ^ ^ ^
NID
Vf \J
N[5)
l ^ W % ^
Nd)
N(3)
NdO)
\J\Mwh^
N(5)
Figure 2:
300 200 100
PPM 15N NMR spectra of riboflavin-binding apoprotein from egg white reconstituted with [l,3,5,10-15N4]7-methy1-10-ribityl-isoalloxazine. Spectrum A: oxidized protein in 100 mM potassium phosphate, pH 6.4. Spectrum B and C: two-electron reduced protein in 100 mM potassium phosphate, pH 6.3 and 100 mM sodium borate, pH 9.0, respectively.
It has been observed that the pKa value of the deprotonation of N(3)H of
protein-bound riboflavin is increased (Miura et al., 1983) as compared to that
of free riboflavin. It is suggested that the weak hydrogen bond to 0(4a) of the
- 103 -
protein-bound r i bo f lav in is responsible for the increased pK value. a
The chemical shifts due to N(10) and C(10a) can be used as monitors with
regard to the polarization of the isoalloxazine ring and the configuration of
the N(10) atom. The N(10) resonance position is shifted even further downfield
than in free FMN. The same holds for C(10a), although to a lesser extent.
This clearly demonstrates that N(10) is forced into the molecular plane and
the isoalloxazine ring is strongly polarized. These results suggest that ribo
flavin interacts strongly with the apoprotein and that the neighborhood of the
N(10) atom and the N(10) side chain of riboflavin are strongly involved in the
interaction. This interpretation is in accord with binding studies of flavin
analogues (Choi and McCormick, 1980).
1Nv10a£Nv>-C, a . R = C H 2 - (CH0H)3 -CH20P0 3 H 2
b: R=CH2-(CH0C0CH3)3-CH20C0CH3
Figure 3: Structure of flavin derivatives enriched with 15N at positions 1,3,5, and 10. R = ribityl, tetraacetylribityl and ribityl-5'-monophosphate.
15
The N NMR data at lower pH values show that polarization or hydrogen bond
ing is hardly altered. The resonance line of N(3) is broadened however (line-
width ^20 Hz) at pH 6.2 compared to that at higher pH values. This broadening
is absent if H,,0 is used as the solvent. This strongly indicates that we are
dealing with a hydrogen exchange reaction, which is slower at low than at high
pH values, in agreement with base catalyzed hydrogen exchange.
- 104 -
§ty^iê§_2D_tb?_£wo-e1ectron_reduced_state
The dissociaton constant of RBPY with reduced riboflavin is about 1 yM
(Scheller et a K , 1979). Consequently, our samples investigated contain about
2% riboflavin free in solution. The dissociation constant of RBPW with ribo
flavin is about 1000 times higher in the reduced than in the oxidized state
(Blankenhorn, 1978), i.e. 8% of free riboflavin is calculated to be present in
our samples. Our chemical shift data will therefore mainly reflect the complex
13 between RBP and riboflavin. Spectra B and C of Figure 1 show C NMR spectra
15 for RBPY at different pH values. Spectra B and C of Figure 2 show N NMR spectra
of RBPW. The chemical shift values are summarized in Table II. The data for
FMNHp and FMNH in water, and TARFH~ in chloroform (cf. also accompanying paper,
Moonen et al_., 1984) are included for comparison.
A comparison of the data for RBPY and RBPW shows that in the reduced state
the differences in the chemical shift values are also within the experimental
error. From this we conclude, in agreement with the results on the oxidized
state, that the binding site for reduced riboflavin is similar in RBPY and RBPW.
Again our analysis concerns consequently both proteins.
In reduced free FMN the deprotonation of N(1)H occurs with a pK of about a
6.6 (Dudley et al_., 1964). Using 1 3C (Van Schagen and Müller, 1981) and 1 5N NMR 15 (Franken et al_., 1983) it has been shown that the chemical shifts due to N(l),
13 13 C(2) and C(10a) reflect this deprotonation clearly. Figure 4 shows the depend-
13 ence of the chemical shift on the pH for RBPW complexed with [2- C]-riboflavin.
The curve and the extreme chemical shift at low and high pH values are identical
with those of reduced FMN in the same buffer, but the pK value is shifted from a
^6.6 to 7.45±0.05. The pK value for RBPY was determined as 7.35±0.05. This sup-a
ports our conclusion drawn above that the binding sites in the two proteins must
be very similar. An explanation for the increased pK, values for the complexes a
as compared with reduced f ree FMN w i l l be offered below.
The chemical s h i f t of C(2) i n the protein-bound, neutral 1,5-dihydroribo-
f l av in is ident ical to that of FMNH«, suggesting that 0(2a) in teracts with
105
Table II. Carbon-13 and Nitrogen-15 chemical shifts of reduced riboflavin water reso
nances is'L10±4 Hz and that of 15N resonances 13±5 Hz, unless otherwise stated.
bound to RBPY or RBPW. For comparison the values for FMNH„ in and TARFHo in chloroform are also included. Linewidth of 13C
Compound Solvent Chemical shifts in ppm
RBPW3)
RBPY3)
FMNH2b)
TARFH2
RBPW3)
RBPY3)
FMNH"b)
RBPWd'e)
RBPWa'd)
RBPYa'd)
MeIMNH2b'd)
MeTARIH2
RBPWa'd)
MeIMNH"b,d)
pH 6.0
pH 6.0
pH 5.0
C2HC13
pH 9.0
pH 9.0
pH 9.0
pH 5.5
pH 6.3
pH 6.3
pH 5.0
C2HC13
pH 9.0
pH 9.0
Carbon Atoms
C(2)
151.1
151.1
151.1
150.6
158.1
158.0
158.2
Nitrogen
N(l)
123.5J)
128.5k)
129.8k)
128.1
116.7
176.0m)
186.9
C(4)
156.5
156.6
158.3
157.0
156.8
157.0
157.7
Atoms
N(3)
148.3
148.21)
148.91)
150.7
145.8
150.3
150.7
C(4a)
102.8
102.8
102.8
105.1
102.8h)
102.4h)
101.4
N(5)
59.1
59.81)
59.91)
60.6
60.7
60.1")
60.6
C(10a)
143.9h)
144.2h)
144.0
137.1
155.0
155.0
155.5
N(10)
89.9
89.9
90.3
87.3
72.2
97.7
97.2
For superscripts a-e, see Table I.
h) Linewidth -\J0 Hz, precision of chemical shift is ±0.6 ppm.
j) Linewidth ^40 Hz, precision of chemical shift is ±0.4 ppm.
k) Linewidth ^80 Hz, precision of chemical shift is ±1.0 ppm.
1) Linewdith ^50 Hz, precision of chemical shift is ±0.8 ppm.
m) Linewdith ^200 Hz, precision of chemical shift is ±2.0 ppm.
n) Doublet with a coupling constant of 78±5 Hz.
106
s: Q_ CL
159
158
157
156
155
154
153
152
c_
PH
Figure 4: The pH-dependence of the 1 3C chemical shift of [2-13C]l,5-dihydroribo-flavin bound to riboflavin-binding apoprotein from egg white. The solid curve has been calculated using a pK value of 7.45.
15 solvent water. No coupling could be observed with FLO as a solvent for the N
resonance of the N(3)H group. This indicates the exposure of this site to the
solvent. The chemical shift of N(3) however, resembles more that of TARFH2 than
that of FMNH2- As in the oxidized molecule a hydrogen bond to 0(4a) is absent,
clearly demonstrated by the upfield position of C(4). The situation is more com
plex with regard to the N(l) atom. The fact that a good titration curve for the
deprotonation of N(1)H could be obtained demonstrates that the protonation/
deprotonation reaction is faster than the chemical shift separation of C(2) in
the two states (Binsch, 1975), which is about 500 Hz. Moreover, no sharp reso
nance line for N(l) could be observed in HLO indicating hydrogen exchange with
the solvent being faster than that expected on the one-bond NH coupling constant.
These data clearly indicate exposure of N(l) to solvent water. On the other hand
the chemical shift correlates better with that of TARFH„ suggesting a weak ex-
- 107 -
posure to solvent water. This apparent discrepancy is explained as follows. The
accompanying paper (Moonen et a K , 1984) shows that N(l) in TARFH? is not of 100%
2 2
sp character and that the sp character of the N(l) and the N(10) atom is increas
ed on polarization of the flavin. Although the N(10) atom thus possesses consider-2
able sp character in the complex, as illustrated by its chemical shift, the N(l) 3
atom shows more sp character than in FMNH„. Steric hindrance by the apo-RBP
could explain this observation, but this explanation is not in agreement with the
easy accessibility to solvent water. It is therefore suggested that the absence
of a hydrogen bond to 0(4a) causes the observed effect. As 0(4a) is an unfavour
able IT electron acceptor in the absence of a hydrogen bond, the derealization of
the TT electrons of N(10) and N(l) is less favoured. The N(10) atom is likely
2
forced into the molecular plane by the apoprotein, which makes sp hybridiza
tion of the N(l) atom unlikely. This could explain why the resonance lines due
to N(l) and also C(10a) are broadened considerably at about pH 6.0 (Table II).
This broadening and the other data suggest, that the N(l) atom in RBP moves in
and out of the molecular plane. The C(10a) atom is probably also influenced by
this motion. At pH 5.5 the linewidth of the N(l) resonance line decreases to
about 50 Hz (Table II), indicating that the frequency of this motion is increased.
The latter result is compatible with the observed swelling of the protein at low 10
pH values (Kumosinski et dj_., 1982). The F NMR data of Miura et jH. (1983) can
be similarly explained. The N(5) atom is also fairly exposed to solvent as no
coupling constant could be detected in hLO.
In the reduced state the degree of bending around the N(5)-N(10) axis
(i.e. the "butterfly" conformation) is expressed directly by the N(5) and N(10)
resonances and indirectly by the C(10a) and C(4a) resonances which are under
the influence of the u electron donating character of the N(5) and N(10) atoms
(Moonen et aj_., 1984). The N(10) atom resonates at about 2 ppm downfield from
that of FMNHp, whereas the N(5) atom is shifted upfield. These results indicate
strongly that the bending is more pronounced at the N(5) than the N(10) position,
as compared to the conformation of free FMNH«. The C(10a) atom resonates at a
position comparable with that in FMNH«, which supports this conclusion. The
- 108 -
absence of a hydrogen bond to 0(4a) disfavours TT electron delocalizaton from
N(10) and N(l) to the 0(4a) position. An increase in the IT electron density is
therefore expected at C(4a). Since the N(l) atom is less sp hybridized than the
N(10) atom however, the cooperative -n electron derealization of the two nitrogen
atoms is decreased as compared to that in FMNH«. The overall apparent effect is
that the resonance due to C(4a) seems to be unaffected.
The conclusions remain roughly the same with regard to the anionic 1,5-di-
hydroriboflavin bound to the protein as given for the neutral reduced form. The
chemical shift of C(2) and the absence of a coupling constant for N(3)H again
indicate the exposure of these centers to solvent water. The resonance position
of C(4) indicates the absence of a hydrogen bond. The upfield shift of the reson-
2 ance due to N(l), as compared with FMNH , indicates that the sp character of
this atom is further decreased as compared to that at the lower pH values. The
considerable linewidth even suggests that the N(l) atom rapidly changes its con
figuration (i.e. configurational averaging). The broadened resonance of C(4a)
might be caused by the configuration averaging of N(l). The results further indi
cate that the configuration of the N(5) and the N(10) atoms are roughly the same
as in the neutral protein-bound flavin molecule. The small downfield shift of
the resonance due to C(4a) as compared with FMNH" supports this interpretation
and indicates that IT electron derealization from the N(l) atom is indeed de
creased-. It can be concluded from these results that most features of the bound
neutral reduced riboflavin are maintained in the negatively charged system. 3
There are two exceptions to this conclusion; i) the sp hybridization of the
N(l) atom is increased; ii) the N(5)H group shows a coupling constant of 78±5 Hz.
The magnitude of this coupling constant clearly shows that the N(5) atom possesses 3
increased sp character, already concluded from its resonance position. The fact
that the coupling constant could only be observed at high pH values indicates
that the protein undergoes a conformational change when the pH of the solution is
increased. This change in conformation shields N(5)H to solvent water at high pH.
The result is compatible with a swelling of the protein at low pH and a shrink
ing at high pH, as proposed recently by Kumosinski et al. (1982).
- 109 -
We return now to the point of the higher pK value for the deprotonation a
of N(l) as compared with reduced free riboflavin. The apparent destabilization of
the negative charge on N(l) might be due to a negative charge in the vicinity
of N(l) in the complex. Another more likely possibility is the absence of a hy
drogen bond to 0(4a). As we have seen in the accompanying paper (Moonen et al.,
1984), the structure of the flavin (especially N(10)) changes upon deprotonation
in such a way that the IT electron derealization of N(l) is also favoured. As
has been indicated the C(4a) and 0(4a) are important centers for this dereali
zation. The derealization to 0(4o) is not favoured in the complex, which pro
bably influences the degree of hybridization of N(l) and thus renders the depro
tonation unfavourable.
The apparently decreased sp character of N(l) in the reduced state might
explain the weaker binding of reduced as compared with oxidized riboflavin. How
ever, at the moment we cannot quantify this effect.
This study has demonstrated that detailed information can be obtained 13 15 from flavoproteins using C and N NMR techniques and isotopically enriched
flavocoenzymes. The information thus obtained is more specific than binding
studies with flavin derivatives (Choi and McCormick, 1980). Some results can
be compared directly with published results such as the solvent accessibility,
which is in excellent agreement with the binding study of flavin derivatives.
Other features such as a strong interaction of 0(4a) and N(10) of riboflavin
with the apoprotein have been suggested from binding studies and are now con
firmed by our results. The data on dynamics and strength of hydrogen bonding
are novel however and describe the characteristics of binding in a very detailed
way.
Finally, in the light of our results we wish to briefly discuss the ques
tion of the function of the protein. It is known (McDonnell et al_., 1951) that
the egg is under anaerobic conditions during the first stage of development.
It might well be that riboflavin is in the reduced state in this stage. The
relative exposure of RBP-bound riboflavin could suggest that reduced flavin
serves as an electron donor for the reduction of the disulfide bonds of ovo-
mucine, a phenomenon which is long known (McDonnell et al., 1951) but still
- 110 -
remains unexplained. Our hypothesis further implies that the mentioned redox
reaction requires the consumption of one proton if the reduced riboflavin is
in the anionic state. This mechanism could also explain the observation that
the pH value of the albumen of a fresh egg is 7.6 and increases to 8.5 within
the first day of development (Carter, 1968). The initial pH value of a fresh
egg is thus above the pK value of the reduced complex, i.e. reduced ribofla-a
vin is predominantly in the anionic state. Thus, our NMR results and the hypo
thesis derived from them, seem to offer an explanation for the observed pH
changes in the developing embryo. The egg yolk is highly structured and pH gra
dients are important for the differentiation of the cells. It is therefore not
unlikely that the redox reactions mentioned are related to the phenomena of pH
gradients.
ACKNOWLEDGEMENT
We are thankful to Mrs. J.C. Toppenberg-Fang for typing the manuscript, to
Mr. M.M. Bouwmans for preparing the figures and to Dr. C.A.H. Rasmussen for
carefully reading the manuscript.
REFERENCES
Allerhand, A. (1979) Methods Enzymol. 61, 458-549. Becvar, J. and Palmer, G. (1982) J.BToTTChem. 257, 5607-5617. Binsch, G. (1975) in: Dynamic Nuclear Magnetic Resonance Spectroscopy (Jackman,
L.M. and Cotton, F.A., Eds.) Academic Press, New York pp. 45-81. Blankenhorn, G., Osuga, D.T., Lee, H.S. and Feeney, R.E. (1975) Biochim.Biophys.
Acta 386, 470-478. Blâni<ënTio7n, G. (1978) Eur.J.Biochem. 82, 155-160. Board, R.G. and Fuller, R. (1974) Biol Rev. 49, 15-49. Bock, K. and Thogerson, H. (1982) in: Annual Report Nuclear Magnetic Resonance
Spectroscopy, Vol. 13 (Webb, G.A., Ed.), Academic Press, London, pp. 1-57. Carter, T.C. (Ed.) (19l>8) in: Egg Quality, Oliver and Boyd, Edinburgh, pp. 26-58. Choi, J. and McCormick, D.B. (1980) ArchTBiochem.Biophys. 204, 41-51. Dudley, K.H., Ehrenberg, A., Hemmerich, P. and Müller, F. 7J"9~64) Helv.Chim.acta
47, 1354-1383. Eweg, J.K., Müller, F. van Dam, H., Terpstra, A. and Oskam, A. (1982)
J.Phys.Chem. 86, 1246-1251. Farrell, H.M.Jr., Mallette, M.F., Buss, E.G. and Clagett, C O . (1969) Biochim.
Biophys.Acta 194, 433-442. Farrell, H.M., Buss, E.G. and Clagett, C O . (1970) Intern.J.Biochem. 1, 168-172. Franken, H.D., Rüterjans, H. and Müller, F. (1983) Eur.J.Biochem., submitted.
- Ill -
Froehlich, J.A., Merrill, A.H.Jr., Clagett, C O . and McCormick, D.B. (1980) Comp.Biochem.Physiol. 66B, 397-401.
Gratwohl, C. and WUthricFTK. (1974) J.Magn.Reson 13, 217-225. Jacubczak, E., Leveau, J.-Y. and Montreuil, J. (196BJ Bull.Soc.Chim.Biol. 50,
2192-2193. Kawabata, M. and Kanamori, M. (1968) Agric.Biol.Chern. 33, 75-79. Kumosinski, T.F., Pessen, H., Farrell, H.M.Jr., (1982)~Ä~rch.Biochem.Biophys.
214, 714-725. Levy, G.C. and Lichter, R.L. (1979) in: Nitrogen-15 Nuclear Magnetic Resonance
Spectroscopy, John Wiley and Sons, New York, pp. 28-107. MacDonnell, L.R., Lineweaver, H. and Feeney, R.E. (1951) Poultry Science 30,
856-863. Matsui, K., Sugimoto, K. and Kasai, S. (1982a) J.Biochem. 91, 469-475. Matsui, K., Sugimoto, K. and Kasai, S. (1982b) J Biochem.' 3T, 1357-1362. Mayhew, S.G. and Ludwig, M.L. (1975) Enzymes, 3rd Ed. 12, pp. 57-118. Miura, R., Kasai, S. Horiike, K., Sugimoto, K., Matsui, K., Yamano, T.
and Miyake, Y. (1983) Biochem.Bi ophys.Res.Comm. 110, 406-411. Moonen, C.T.W. and Müller, F. (1983) Eur.J.Biochem7T33, 463-470. Moonen, C.T.W., Vervoort, J. and Müller, F. (1984), accompanying paper. Müller, F. (1972) Z.Naturforsch. 27b, 1023-1026. Müller, F., Vervoort, J., Lee, J., Horowitz, M. and Carreira, L.A. (1983)
J.Raman.Spectr. 14, 106-117. Murthy, U.S., Sreel<rishna, K. and Adiga, P.R. (1979) Anal .Biochem. 92, 345-350. Nishikimi, M. and Kyogoku, Y. (1973) J .Biochem. 1_, 1233-1242. Nishina, Y. (1977) Osaka Daigaku Igahu Zasshi 29, 261-269. Ostrowski, W. and Krawczyk, A. (1963) Acta Chem.Scand. 17, S241-S245. Ostrowski, W., Zak, Z. and Krawczyk, A. (1968) Acta Biocïïim.Pol. 15, 241-259. Osuga, D.T. and Feeney, R.E. (1968) Arch.Biochem.Biophys. 124, 560~r574. Scheller, F., Strarad, G., Neumann, B., Kuhm, M. and Ostrowski (1979)
J.Electroanal.Chem. 104, 117-122. Steczko, J. and OstrowsFT, W. (1975) Biochim.Biophys.Acta 393, 253-266. Van Schagen, C G . and Müller, F. (1981) Eur.J.BiochemT"T2"07~3~3-39. Winter, W.P., Buss, E.G., Clagett, C O . and Boucher, R/VT~(1967)
Comp.Biochem.Physiol. 22, 897-906. Witanowski, M., Stefania¥7 L. and Webb, G.A. (1981) in: Annual Report Nuclear
Magnetic Resonance Spectroscopy (Webb, G.A., Ed.) Vol. IIB, pp. 1-493. Yagi, K., Ohishi, N., Takai, A., Kawano, K. and Kyogoku, Y7"( 1976) in:
Flavins and Flavoproteins (Singer, T.P., Ed.) Elsevier Scientific Publishing Company, Amsterdam, pp. 775-781.
Zak, Z., Ostrowski, W., Steczko, J., Miroslawa, W., Gizler, H. and Morawiecki, A. (1972) Acta Biochim.Pol. 19, 307-323.
- 112 -
Chapter 9
The different classes of flavoproteins: Miscellaneous NMR results
and mechanistic implications
9.1. Introduction
In chapter 1 a classification of the different flavoproteins is given. In
this chapter NMR results of a particular flavoprotein belonging to one of these
classes of flavoproteins are discussed, i.e. the flavodoxins, the dehydrogenases
and the hydroxylases. Flavodoxins, especially that from Megasphaera elsdenii
has been discussed already in previous chapters and will be considered further
below, where attention is paid to the redox potential (9.2). Lipoamide dehydro
genase from Azotobacter vinelandii has been chosen as a representative of the
dehydrogenases (9.3) and parahydroxybenzoate hydroxylase from Pseudomonas fluorescens
as a representative of the hydroxylases (9.4.).
9.2. Modulation of the redox potential in flavodoxin.
In the foregoing chapters various results on M.elsdenii flavodoxin have
been discussed, i.e. the mobility of the prosthetic group FMN, the structure
of protein-bound FMN and the protein in the different redox states and the
activation barriers in the redox transitons. In this paragraph attention is
paid to the modulation of the redox potential of FMN by the apoprotein. In Table
1 the redox potentials of FMN and some flavodoxins are summarized. Although
the redox potentials among the flavodoxins show only minor differences, a dras
tic change is observed with regard to free FMN, amounting to about a 300 mV
decrease for the E, and +150 mV for the E 2 value. Smith et al_. (1977) pointed
out that the E~ value is difficult to rationalize as the protein undergoes
distinct conformational changes by the transition from the oxidized to the semi-
quinone state.
- 113 -
Table 1. Redox potentials of some flavodoxins and free FMN. E, represents
the redox potential of the transition between the semiquinone and
the hydroquinone state, E 2 represents the redox potential of the
transition between the oxidized and the semiquinone state. For
the redox potentials of the flavodoxins, see Mayhew and Ludwig
(1975) and references therein. For the redox potentials of FMN,
see Anderson (1983).
Redox potentials
Compound pH E-^mV) E^^v)
M.elsdenii flavodoxin 8 -375, -413a) -175
Clostridium MP flavodoxin 8 -408 -152
D.vulgaris flavodoxin 7.8 -440 -150
FMN -124 -314
a) Calculated value for I = 100 mM, see text.
However, the situation is much simpler for the biologically relevant redox
transition between the semiquinone and the hydroquinone state, as crystal-
lographic data on Clostridium MP flavodoxin (Ludwig et a K , 1977) and NMR
data on M.elsdenii flavodoxin (Moonen and Müller, 1982, 1984) have shown
that the protein structure must be extremely similar in these two states.
Moreover, crystallographic data (Smith et a^L, 1977, Ludwig et aJL. 1976) and
NMR data (Van Schagen and Müller, 1981; Franken et a]_., 1983) have shown
that the isoalloxazine ring is almost coplanar in both redox states. An iden
tical structure with respect to the flavin is also observed in Desulfovibrio
vulgaris flavodoxin (Watenpaugh et aj_., 1976). This has led Simondsen and
Tollin (1980) and Tauscher et a K (1973) to propose that the low E, value
in flavodoxins is a result of the coplanar structure of flavin which is
forced onto the flavin by the apoprotein. At that time it was thought that a
- 114 -
planar conformation of reduced flavin is energetically not favored. However,
it has been shown recently (Moonen et al_., 1984a) that polarized free FMNH" 2
is hardly bent, i.e. the N(10) is almost fully sp hybridized, whereas N(5)
2 3 is approximately 70% sp and 30% sp hybridized. Moreover, the activation
energy of the ring inversion (or "butterfly" motion) must be very low as deduced
13 from dynamic C NMR results (Moonen et al^, 1984b), theoretical calculations
(Dixon et al_., 1979) and 1 3C and 1 5N NMR data of model compounds (for
references see Moonen et a K , 1984a and 1984b). Therefore the observed small
bending angle in flavodoxin cannot cause the drastically decreased redox po
tential. The obvious alternative origin until now not considered, is the influence
of the negative charge at N(l) in the hydroquinone state, which has been
demonstrated to exist by NMR results (Van Schagen and Muller, 1980; Franken
et al_., 1983). It should be mentioned that the neutral semiquinone state is present
in flavodoxins. It should,however, be possible to quantify the effect of the
negative charge on N(l) on the redox transition E,, as (from NMR and crystal-
lographic results) charges are located in the vicinity of the flavin .
For example, it has been shown that the phosphate group of FMN in M.elsdenii
flavodoxin is in the dianionic form regardless of the redox state (Moonen and
Müller, 1982). From this study, the distance to the N(l) atom is also known i.e.
about 0.8 nm. Moreover, this flavodoxin contains a large net negative charge (-17),
contributing also to repulsion of the charge at N(l) of flavin. A third effect
is the monopole-dipole interaction between the latter charge and the dipole of
the protein. The relevant equations for the quantification of the charge inter
actions were developed recently by Van Leeuwen (1983). In the following the dif
ferent contributions to the E, value will be discussed.
1. The monopole-monopole interaction of the negative charge at N(l) of flavin
and the dianionic phosphate group: The distance between the charges is 0.8 nm
(Moonen and Muller, 1982). No countercharges are near the N(l) atom and the phos
phate group of flavin, or in the region between the charges as evidenced by
- 115 -
crystal!ographic results on the related Clostridium MP flavodoxin (Ludwig
et a K , 1976). Both the phosphate group and the isoalloxazine ring are tightly
bound and buried in the protein (Moonen and Müller, 1982). The charges are thus
separated by a rather low dielectric constant. A reasonable estimation for
the dielectric constant is 20. This results in a repulsion energy of 7.2 kT or
180 mV.
2. The monopole-monopole interaction of the negative charge of the N(l) atom of
flavin and the overall net negative charge of the flavodoxin.
M.elsdenii flavodoxin possesses a net negative charge of -15 (excluded
the phosphate group). Upon comparison to the related flavodoxin from Clostridium
MP several charged residues appear to be invariant in the primary sequence
(Mayhew and Ludwig, 1975). Crystallographic data show that all charges (ex
cept that of the phosphate moiety) are exposed to solvent. The charge inter
action is thus sensitive to the ionic strength of the solvent. At a very low
ionic strength (I = 0) a charge repulsion of 160 mV is calculated, for I = 100
mM the repulsion is 57 mV.
3. The monopole-dipole interaction of the negative charge at N(l) and the overall
dipole moment of the protein:
Recently, the dipole moment of Clostridium MP flavodoxin has been calcu
lated (Van Leeuwen, unpublished results). The dipole moment is 450 Debye with
the negative pole pointing towards the isoalloxazine moiety. From these data
a repulsion energy of 52 mV is calculated.
From the calculations presented above it follows that only interaction 2
is dependent on the ionic strength. Thus, using I = 100 mM the total repulsion
is 289 mV and using I = 0 mM the total repulsion is 392 mV. This result strongly
suggests that the negative charge at N(l) in the hydroquinone state causes the
dramatic decrease in the redox potential owing to strong repulsion. The close
resemblance of the actual and calculated values of the redox potentials shows that
the almost planar structure of the fully reduced isoalloxazine ring in flavodoxin
- 116 -
is not the cause of the decrease of the redox potential.
The analyses given above clearly show that redox potential modulation can
be easily achieved by charge interactions. As will be shown in subsequent paragraphs
the negative charge at N(l) of flavin is found in all flavoproteins investigated
so far in our laboratory (i.e. luciferase from Vibrio harveyi, flavodoxin from
M.elsdenii and D.vulgaris, lipoamide dehydrogenase from A.vinelandii and parahydroxy
benzoate hydroxylase from P.fluorescens). Thus charge interactions are in all
these cases potential mechanisms for redox potential modulation. It should be
noted that charged substrates might play also a significant role in this modu
lation.
9.3. Lipoamide dehydrogenase
9.3.1. Lipoamide dehydrogenase (NADH: lipoamide oxidoreductase, EC 1.6.4.3)
catalyzes the reversible reaction
NAD+ + dihydrolipoamide *» lipoamide + NADH + H .
The enzyme contains two redox centers, disulfide and FAD. Consequently, the
enzyme can be fully reduced by four electrons, which species is thought to be
catalytically inactive. The reaction mechanism, proposed by Massey and Veeger
(1961) shows the essential role of the two-electron reduced enzyme (EH„), which
has later been confirmed by many investigators (for a review, see Williams (1976)).
The first step in the reaction is the oxidation of dihydrolipoamide by the disulfide
thereby forming the EH,, species. In this form of the enzyme the electrons are
shared by the disulfide group and FAD. This species is characterized by a
'charge-transfer' band at 530 nm. Although several proposals have been made for
the electronic structure of EH«, (Palmer and Massey, 1968; Searls et al_., 1961;
Massey et a]_., 1960), recently Wilkinson and Williams (1979) clearly showed
for the E.coli enzyme that EH„ is in fact a mixture of distinct species. The
second step in the reaction is the reduction of NAD by EH„ yielding the oxidized
enzyme and NADH. Several kinetic data are in agreement with the model, but others
remain difficult to explain (for a review see Williams, 1976). Among these especially
- 117 -
the requirement for NAD+ in the NADH oxidation by the enzyme and the possible
significance of a second binding site for NAD are mentioned. Matthews et al.,
(1979) suggested that NAD functions not only as an acceptor but also as an
effector which influences flavin-disulfide interaction in EH~. In this study
13 specifically C labeled FAD is used to investigate the interactions between the
apoflavoprotein of Azotobacter vinelandii and FAD in the oxidized, the EHL and
the 4 electron reduced state (EhL). Moreover, the electron distribution in EH~
is investigated resulting in additional mechanistic information.
9.3.2 Materials and Methods
Lipoamide dehydrogenase (Lipdh) from Azotobacter vinelandii was isolated
and purified from the a ketoglutarate dehydrogenase complex as described by
Bosma (1984). Preparation of the apoprotein was performed as follows. Puri
fied enzyme (140 mg) was loaded on a phenylsepharose column (Pharmacia) using
a 50 mM potassium phosphate buffer which contained 30% ammonium-sulphate, 1 M
KBr and 1 mM EDTA (final pH 6.5). Subsequently, FAD was removed from the pro
tein by washing the column with 10 ml of 50 mM potassium phosphate buffer, con
taining 30% ammoniumsulphate, 1 M KBr and 1 mM EDTA (final pH 4.0).
Subsequently the column was washed with 15 ml of the same mixture omitting KBr.
13 Recombination with C labeled FAD was performed on the column immediately after
the preparation of apoprotein using 2 ml of a solution containing 5 mM FAD,
50 mM potassium phosphate, 30% ammonium sulphate, 1 mM EDTA (final pH 6.5).
The column was then washed with the same mixture, from which FAD was omitted,
for 4 h at room temperature and subsequently for 16 h at 4°C. The reconstituted
protein was eluted from the column with a mixture of 10 mM potassium phosphate
containing 1 mM EDTA and 50% ethyl eneglycol (final pH 6.9). The first 3 ml of
the eluate contained about 75% of the total protein reconstituted. The re
mainder eluted in a rather diffuse band of about 50 ml. The reconstituted pro
tein was collected and subsequently dialyzed and concentrated using an ultra-
- 118 -
filtration procedure (Amicon). The recovery of reconstituted protein was higher
than 80% of the starting material. The specific activity and absorption spectrum
of the reconstituted protein were identical with those of the starting material.
This procedure appeared to be superior to the procedure described by Van den
Broek (1971). 1-3
C NMR measurements were performed with a Bruker CXP 300 spectrometer
operating at 75.5 MHz. 10 min Wilmad precision tubes were used. Samples contained
1.6 ml solution consisting of 1.2 mM protein and 50 mM potassium phosphate
buffer (pH 7.0). 10% D„0 was added for the stabilization of the magnetic field
and 3 yl dioxan was used as the internal reference. 30° pulses and a repetition
time of 0.6-0.8 s were applied. Quadrature detection and quadrature phase
cycling were used. Temperature was kept constant at 26 C. Anaerobiosis was
achieved by carefully flushing the sample with argon for 20 min. Reduction was
performed with dithionite (Mayhew, 1978). Oxygen-free samples were sealed 13 with a serum cap. C enrichment of the isoalloxazine ring was performed
as described by Van Schagen and Müller (1980). FAD was synthesized and purified
as described by Cramer and Neunhoeffer (1962).
9.3.3. Results
9.3.3.1. The oxidized state.
13 In Fig. 1 the C spectrum is shown of oxidized lipoamide dehydrogenase,
enriched with 1 3C (>90%) at the position C(4) and C(10a) of FAD. The resonance
due to C(4) is overlapping with a natural abundance resonance at 158.2 ppm,
probably representing the C atom of arginine residues. Where necessary, dif-
13 ference spectra between natural abundance spectra and C spectra of samples
13 containing C enriched flavin were performed to identify the resonances due
to the prosthetic group. The chemical shift due to C(2), C(4), C(4a) and C(10a)
atoms in oxidized and 4 electron reduced lipoamide dehydrogenase are summarized
13 in Table 2. In a previous C NMR study on the interaction between riboflavin
- 119 -
13 15 and riboflavin binding protein it was shown that the combination of C and N
NMR data allow a detailed description of the structure of the flavin and specific
hydrogen bonds between FAD and the apoprotein (Moonen et a]_., 1984). Therefore
180 150 120 90
PPM (6)
60 30
13 Figure 1. C NMR spectrum of oxidized lipoamide dehydrogenase reconstituted with specifically 1 3C enriched (>90%) FAD at position C(4) and C(10a). The position of these atoms is indicated in the spectrum. Chemical shifts are reported relative to TMS. As the internal reference dioxan was used (indicated by the arrow at 67.8 ppm). Data acquisition lasted approximately 12h. Sample contained 1.2 mM enzyme (on the basis of M = 56 000) in 50 mM potassium phosphate, pH 7.0).
13 one must be cautious in this study because only C data are available. In
Table 2 for comparison the chemical shifts of the isoalloxazine ring in water
(i.e. monomeric FMN) and in chloroform (i.e. tetraacetylriboflavin) are added.
In oxidized lipoamide dehydrogenase both resonances due to the C(2) and the C(4)
of flavin are downfield shifted as compared with TARF in chloroform, and are
almost identical with those of FMN. This clearly demonstrates that both carbonyl
groups are polarized indicating hydrogen bond formation, although slightly
- 120 -
weaker than for FMN in water. The downfield shifted resonance due to
C(4a) indicates that N(10) is a weak electron donor and that it is probable
somewhat out of the molecular plane. For free isoalloxazine this occurs also
in a medium with a relatively low dielectric constant (Moonen et al_., 1984). The
downfield shifted resonance of C(10a) can be caused by two effects i) by mesomeric
structures via the polarization of 0(4a) and 0(2a) ii) by the absence of a
15 hydrogen bond to the pyridine-like N(l) atom, without N NMR data it is not
possible to determine the actual background of this shift. However, summarizing
the general picture of the oxidized FAD bound to lipoamide dehydrogenase,
it is concluded that both carbonyl groups are fairly polarized, but the
polarization seems not to extend to the N(10) nucleus, probably due to a
microenvironment of a relatively low dielectric constant. It has already been
concluded from absorption data that the isoalloxazine ring is in a hydrophobic
environment (Williams, 1976). This is in rough agreement with our data, but
it should be mentioned that despite the hydrophobic milieu the carbonyl groups
are polarized.
9.3.3.2. The 4-electron reduced state.
The chemical shifts of C(2) and C(10a) of the 4-electron reduced lipoamide
dehydrogenase (EH.) clearly indicate that the N(l) atom is ionic in 50 mM
potassium phosphate buffer, pH 7.0. From the chemical shifts of C(2) and C(4)
it can be concluded that both carbonyl groups are strongly polarized. In addition,
the resonance line of C(4a) is strongly upfield shifted as compared with FMNH",
showing that the N(10) atom is a strong -n electron donor and thus must have al-
2 most full sp character. The resonances of C(10a) is somewhat upfield shifted
as compared to FMNH . This suggests that the degree of sp hybridization of the N(5)
13 atom is slightly increased compared with FMNH . Thus, the combined C NMR data
of FADH in EH, show that both carbonyl functions are polarized and that the
isoalloxazine ring is probably almost planar. The latter conclusion suggests that
the reduced bound isoalloxazine ring is in a hydrophilic milieu in contrast to
- 121 -
13
Table 2. Chemical shift of C labelled atoms of FAD bound to lipoamide dehy
drogenase (50 mM, potassium phosphate, pH 7.0). For comparison the chemical
shifts of FMN in water and TARF in chloroform are added, a) lipdh is
lipoamide dehydrogenase, b) FMN is flavin mononucleotide, c) TARF is
tetraacetylriboflavin. The accuracy of the reported chemical shifts
is ± 0.2 ppm.
lip.dha (oxidized)
FMN in water
TARF in chloroform
lip.dha(2e~ reduced)
lip.dha(4e" reduced)
FMNH in water
FMNH2b in water
TARFHp0 in chloroform
C(2)
158.6
159.8
155.2
159.0
159.0
159.4
158.2
151.1
150.6
C(4)
163.1
163.7
159.8
163.6
157.4
157.5
157.7
158.3
157.0
C(4a)
137.8
136.2
135.6
140.5
101.5
98.9
101.4
102.8
105.2
C(10a
152.0
152.1
149.1
152.9
154.3
154.7
155.5
144.0
137.1
FAD in the oxidized complex. Although the data should be regarded with some caution
it is suggested that a conformational change takes place by the 4 electron reduc
tion of the protein rendering the reduced FADH" more accessible to solvent water.
9.3.3.3. The two electron reduced state (EH2)
It is known that the stability of the two electron reduced state (EH2) against
overreduction is dependent on the source of the protein (Williams, 1976).
Recently, it was suggested that also the composition of EhL (i.e. the different
possibilities of accommodating the reduction equivalents) is dependent on the
source of isolation (Wilkinson and Williams (1979)). The EH2 state of the enzyme
from A.vinelandii resembles in stability and absorption spectrum very much the
- 122 -
EH? of the protein from pig heart (De Kok et a]_., 1983, De Kok, personal
communication). Therefore, we will compare our results mainly with the results
obtained with pig heart lipoamide dehydrogenase.
As with the pig heart enzyme, the red intermediate EH2 is formed by a stoichio
metric titration of the oxidized enzyme with dithionite. The same species is slowly
formed in the presence of excess NADPH and is extremely stable under anaerobic
conditions (De Kok et al_., 1983). Both methods result in similar NMR spectra. Except
that the spectrum obtained by NADPH reduction shows also the natural abundance
resonances of the pyridine nucleotide (data not shown). In Fig.2 the low field
13 part of the C NMR spectra of EH2 are shown. The spectra were obtained with FAD
13 enriched with C at the C(4) and C(10a) position (upper spectrum) and at the C(2)
and C(4a) position (lower spectrum). In the upper spectrum four resonances due
to the carbon atoms of flavin can be clearly seen, although only two positions are
enriched. It should be noted that none of the resonance positions is equal to
the resonance positions of either those of the oxidized or 4 electron reduced
enzyme. This clearly indicates that one deals with two distinct species in the
EH« state. The peaks at 163.6 and 157.4 ppm appear at a position, where C(4)
is about expected in the oxidized and reduced free isoalloxazine, respectively
(Moonen et al_., 1984). The peaks at 152.9 and 154.3 ppm are at positions where
about the resonance of C(10a) is expected in the oxidized and reduced isoalloxa
zine, respectively. These results show that EH2 consists of a mixture of oxidized
and reduced isoalloxazine. This is fully confirmed by the lower spectrum, where
the resonances at 140.5 and 101.5 ppm resemble the C(4a) position in oxidized and re
duced isoalloxazine respectively. Only C(2) does not show a splitting, which is
not unexpected because the chemical shift of C(2) hardly differs in isoalloxazine
upon reduction. This result closely resembles the result of Wilkinson and Wil
liams (1979), who showed, using stopped flow measurements on the E.coli enzyme,
that EH2 consists of a mixture of species in which FAD and disulfide share
the electrons differently. Thus EH2 of A.vinelandii consists of a species of re-
- 123
180 160 140 120 100
PPM (s)
Figure 2. Low field part of the 1 3C NMR spectra of the two electron reduced (red intermediate) lipoamide dehydrogenase. Reduction was performed by a stoichiometric titration with sodium dithionite (cf. Materials and Methods) Upper spectrum enzyme enriched at the positions C(4) and C(10a). Lower spectrum, enzyme enriched at the positions C(2) and C(4a). For other details see text and Caption of Fig. 1.
- 124 -
duced FAD and oxidized disulfide (EH2(A)), and a species of oxidized FAD and
reduced disulfide (EH2(B)). To accumulate the two spectra shown in Fig. 2 about
20 h were needed during which time the sample remained anaerobic. The typical
red colour of the EH„ before and after the measurement was the same. No over
reduction nor oxidation occurred during the time course of the experiment as
within the signal to noise ratio of the experiment no fully oxidized nor 4 electron
reduced enzyme could be detected. The combined intensity of the resonances of the
two EH„ species is equal (±20%) to the oxidized or EH. state and indicates that by
approximation only the two mentioned EH„ species constitute the two-electron reduced
state of the enzyme. The resonances are summarized in Table 2. The EH2(B) species
shows small but distinct differences with the fully oxidized enzyme. This indicates
that either a conformational change has occurred or the reduced disulfide influences
the chemical shift position in the EH? species. The same can be stated for the dif
ference between the four-electron reduced enzyme and the EH?(A) species. In the
following the electronic structure of the two distinct EH2 species will be dis
cussed. The differences in chemical shifts due to C(4a) and C(10a) of the oxidized
isoalloxazine ring in the fully oxidized enzyme and in the EH2(B) species are
remarkable. In the EH2(B) species both resonances show a strong downfield shift,
which cannot be explained solely by hydrogen bond formation or polarization by
a high dielectric medium. In fact the data are consistent only with a negative
charge, very close to these atoms. Crystallographic results on glutathione reductase
(a flavoprotein with a very similar reaction mechanism and physical properties)
show that one thiol of the disulfide moiety is extremely close to C(4a) (Schulz
et aj_., 1978). Therefore, the obvious candidate for the negative charge is the
reduced disulfide bond, in agreement with published results (Williams, 1976).
The small distance between the thiolate anion and the C(4a) and C(10a) centers of
flavin might well cause the 'charge transfer' band in EH?.
The EH2 (A) species in which the isoalloxazine is reduced, resembles
FMNH" in water, with the exception that C(2) is slightly downfield and
- 125 -
C(10a) slightly upfield shifted. This shows that the TT electron density is enhanced
at the C(10a) position at the cost of the IT electron density at C(2). The exact
reason for this is not yet known. It should be noted, however, that on the basis
13 of the C NMR results it cannot be fully excluded that the EH„ absorption
band at 530 nm results from a strong interaction between reduced isoalloxazine and
oxidized disulfide, in contrast to the idea that this band originates from an
interaction between the thiolate anion and the oxidized isoalloxazine ring.
9.3.4. Mechanistic implication
9.3.4.1. Charge interaction and redox potentials.
The results demonstrate that EH2 consists of two different species, one
with the FAD reduced to FADH~ and the disulfide oxidized, (EH2(A)) the other
species with FAD oxidized and the disulfide reduced to SH and S~ (EH2(B)). In both
species the two redox centers are very close to each other. The fact that the two
species are roughly equally populated (±10%) clearly show that the redox potentials
of the couple E /EH2 are almost the same for both species of EhL. The redox po
tential of the pig heart enzyme for the couple E /EH? is -280 mV (pH 7.0) and
-346 mV for EH2/EH« (Matthews and Williamson, 1976). As stability and kinetics
are rougly the same for the enzyme from pig heart and A.vinelandii it is probably
safe to assume that the redox potentials for the latter enzyme are the same.
As the redox potential E /EH2 is the same for both EH2 species and 66 mV more po
sitive than for the couple EH2/EH. it must be concluded that for both the EH2(A)
and the EH2(B) species the reduction of the disulfide or FAD to form EH.
from EH2 requires more energy than the reduction of the same centers in the
transition E /EhL. Thus, if the disulfide is reduced, then the reduction of FAD
is energetically less favorable than in the case where the disulfide is oxidized.
If, on the other hand, FAD is reduced then the reduction of the disulfide is
energetically less favorable than if FAD is oxidized. Recalling the discussion
on flavodoxin (see 9.2), the obvious reason is a charge repulsion occurring
between FADH" and S~ in the EH. state. This conclusion is in full
- 126 -
agreement with the early experiments of Massey and Veeger (1961) using arsenite
incubation of the enzyme. These results show that both thiols originating from
the reduction of the disulfide can be covalently linked with arsenite.
Subsequently FAD is easily reduced by NADH, in contrast to untreated EH2, indi
cating that the redox potential of the FAD center is increased by arsenite ad
dition. It is clear that the formation of a complex between arsenite and both thiols
results in a neutral complex, and as a consequence, no charge repulsion exists
in EH,. Also methylation of one or both of the thiols results in an increased
redox potential of EH^/EH« in the A.vinelandii enzyme (De Kok, personal communication),
It should be noted that the charge repulsion in EH» is less than the charge re
pulsion between phosphate and the charge at N(l) in fully reduced flavodoxins, al
though the distance between the charges in the latter species is probably larger.
The reason might be a partial neutralisation of one of the charges in EH«. There
is indeed some indication that the thiolate anion is stabilized by a base
(Matthews and Williamson, 1976).
9.3.5. The intramolecular electron transfer and the role of NAD
To reiterate the reaction cycle according to Wilkinson and Williams (1979):
In the first step dihydrolipoamide becomes oxidized and the electrons enter
the active site of the enzyme by reduction of the disulfide. The second step
is an equilibration between the different species of EH~ as shown before. The
third step is the reduction of pyridine nucleotide (NAD ) . This third step is
probably an electron transfer from reduced FAD to NAD+ (Williams, 1976). The
second step is schematically represented in the following scheme:
p FAD r FADH"
- S" -• - S
I L SH L s
- 127 -
This step consists of an intramolecular two-electron transfer from the re
duced disulfide to oxidized FAD. It should be noted that this is an essential
step in the reaction cycle as electrons enter via the disulfide moiety and leave
via the FAD moiety. The kinetics of the intramolecular two-electron transfer has
not been implicated in the original kinetic models for the overall reaction
and might explain the fact that no kinetic model could sufficiently explain 1-3
the kinetic data (Williams, 1976). The C NMR data actually show the importance
of the kinetics of the intramolecular electron transfer as under the conditions
of the experiments the lifetime of the electron pair in each EH„ species is
» 10 ms. This conclusion can be directly derived from the fact that the two EH-
species are separately observed in the NMR spectra, although the separation of
the C(10a) resonances in the two EH~ species is about 100Hz. Moreover, the
linewidth of both resonances hardly differs from linewidths observed in the
oxidized state indicating that the exchange frequency hardly causes a linebroadening
effect. A broadening > 5Hz would be obvious from the NMR spectra. The exchange
rate is thus smaller than 5 s" . Thus at 26°C in 50 mM potassium phosphate
(pH 7.0) the lifetime x of each species is > 200 ms. This shows that the kinetics
of this step is likely of great importance in the overall reaction. The low
exchange rate might also be the reason for the slow comproportionation reaction
(Wilkinson and Williams, 1979).
Stopped flow experiments have shown that the oxidation of EH~ by NAD+ is
complete within 3 ms (Massey et al_., 1960). It was concluded that the reaction
rate is faster than 3300 s~ at saturating substrate concentration. It should
be noted that the intramolecular electron transfer, depicted in the above
scheme, is necessary for the completion of the oxidation of EH2 by NAD+. The
combined kinetic data strongly indicate that NAD not only functions as a
substrate in the overall reaction, but in addition increases considerably
the rate of intramolecular electron transfer. NAD functions thus also as an
effector. Matthews et aj_. (1979) suggested already that oxidized pyridine nu
cleotide functions as an effector for the disulfide-flavin interaction.
- 128 -
The C NMR experiments have shown that this suggestion must be correct. These
results offer a good explanation for the kinetics of the reversed reaction.
It has been shown by Massey and Veeger (1961) that for the oxidation of NADH
by lipoamide dehydrogenase the presence of NAD is required. Our data in
dicate that without NAD the slow intramolecular electron transfer limits the
reduction of lipoamide.
9.3.6. Summary
13 C NMR data of lipoamide dehydrogenase reconstituted with FAD, enriched
at the positions C(2), C(4), C(4a) and C(10a), allows a rough determination of
polarization of the isoalloxazine moiety in the oxidized enzyme (E ) , the two-
electron reduced enzyme (EH2) and the four-electron reduced enzyme (EH,). The
results are consistent with a disulfide, located close to the isoalloxazine ring,
which influences (dependent on the redox state) the chemical shifts and redox
potential of FAD. It is suggested that charge interactions play an important
role in the redox potentials. If FAD is reduced, it is present in the anionic
form. It is shown that EHp consists of two species, one in which FAD is oxi
dized a nd the disulfide reduced. In the other one FAD is reduced and the disul
fide oxidized. The redox potential of E0X/EH2 is nearly the same for both
EH2 species. The intramolecular two-electron transfer rate is smaller than
5 s" under the conditions of the experiments. Comparison with reported kinetic
data showed that NAD+ acts as an effector in this intermolecular transfer
process.
9.4. P-hydroxybenzoate hydroxylase
9.4.1. Introduction
p-Hydroxybenzoate hydroxylase (pHBH) EC 1.14.13.2) from Pseudomonas fluorescens
is an NADPH dependent, FAD containing monooxygenase catalyzing the hydroxylation
of p-hydroxybenzoate (pHB) to form 3,4-dihydroxybenzoate in the presence of
molecular oxygen:
129
HC\
HO-fV-COCr+NADPH + H++0 * H O - / V c O O " + NADP++ ^ 0
Two ternary complexes are involved in the reaction cycle; i.e. the formation
of the E(FAD)-pHB-NADPH complex and after reduction of E(FAD)-pHB the formation
of E(FADH?)-0?-pHB. The latter complex is involved in the activation of molecular
oxygen to hydroxylate the substrate. Subsequently the product is released and
the next catalytic cycle can be executed.(Husain and Massey, 1979).
The binding of pHB to the oxidized state to form a 1:1 complex causes
large, characteristic spectral perturbations and fluorescence quenching. The
reduction of bound FAD is dramatically enhanced upon substrate binding (Husain
and Massey, 1979). The structure of the pHBH-pHB comlex is known by crystal-
lographic results (Wierenga et a\_., 1979). Significant progress has further been
obtained on i) the involvement of some "essential" aminoacids (Wijnands and
Muller, 1982; Shoun and Beppu, 1982) ii) the quaternary structure (Müller,
et a K , 1979). iii) the amino acid sequence (Weyer et al_., 1982) and the binding
of FAD (Müller and Van Berkel, 1983).
The exact sequence of processes after the formation of the second ternary
complex remains still uncertain, although some progress has been booked (Entsch
et al_., 1980; Claiborne et al_., 1982; Entsch et al_., 1976; Bruice, 1980; and
references therein).
At the moment it is generally accepted that the activation of oxygen
proceeds via a flavin C(4a)-hydroperoxide intermediate. Its formation is not
fully established. In addition in principle the reaction between singlet
flavin and triplet oxygen is spin forbidden. The oxygen transfer reaction
to the substrate is unclear. While Entsch et al_. (1976) proposed the trans
fer of one oxygen atom, Kemal and Bruice (1976) proposed a mechanism in which
- 130 -
a substrate-peroxide is formed intermediately.
pHBH from other sources than P.fluorescens appears to be similar with
respect to the reaction mechanism and molecular properties.
9.4.2. Materials and Methods
p-hydrobenzoate hydroxylase from P.fluorescens was isolated and purified
as described by Müller et al. (1979). Preparation of apoprotein and reconstitution
was performed as described by Müller and Van Berkel (1983). Enrichment of FAD
13 by C was achieved as described in paragraph 9.3.2.
Samples consisted of 0.6 mM pHBH in 50 mM potassium phosphate (pH 7.0)
10% FLO served as a lock for the spectrometer. Dioxan (3yl) was used as an
internal reference. 10 mm NMR precision tubes were used. The sample volume was
1.6 ml. Spectra were taken on a Bruker CXP 300 spectrometer operating at 75.5 MHz.
Broadband decoupling (0.5 Watt) was used. 30° pulses and a repetition time of
0.6-0.8s were used. Quadrature detection and quadrature phase cycling were ap
plied. Temperature was kept constant at 20°C. For the reduction and ana-
erobiosis procedure see 9.3.2.
9.4.3. Results and Discussion
13
Fig. 3 show* a series of C NMR spectra of pHBH in dependence on the re
dox state and the absence or the presence of substrate. The resonance frequencies
of the enriched carbon atoms C(4) and C(10a) are easily determined either directly
from the spectra or using difference spectra (see also caption of Fig. 3 ) .
The data are summarized in Table 3. In Fig. 4 similar experiments are
shown using FAD enriched at the position C(2) and C(4a) (Table 3 ) . For
comparison of the chemical shifts with those of free flavin the reader is
referred to Table 2 and to a detailed discussion by Moonen and Müller (1984).
As in the case of lipoamide dehydrogenase, the results have to be regarded
13 with some caution, as they are based on C NMR data only.
131
B
D
*^«^sAyri^»«^**H\idW
180 160 140 120
PPM (5)
100
13 Fig. 3. Low field part of the C NMR spectra of p-hydroxybenzoate hydroxylase.
Samples contained 0.6 mM enzyme in 50 mM potassium phosphate (pH 7.0). Chemical shifts are referred to TMS. Dioxan (3 yl) served as an internal reference. All spectra were accumulated in 8 to 16 hours. A. Natural abundance spectrum B-E, Enzyme reconstituted with FAD containing 1 3C isotope (>90%) at the position C(4) and C(10a), B, oxidized state, C, oxidized state in the presence of 5 mM substrate, D, reduced state, E, reduced state in the presence of 5 mM substrate. Resonances marked with A represent natural abundance intensity of substrate. Resonances marked 0 and • represent the C(4) and C(10a) atom, respectively. In the spectra D and E the resonance due to C(4) is partially obscured by the natural abundance resonance at 158.2, probably representing the C atoms of arginines. The position of C(4) in the spectra D and E was therefore determined by difference spectra between A and D, and between A and E. Data are summarized in Table 3.
- 132 -
I!}?_°xldized_state
Without substrate, the polarization of the isoalloxazine ring resembles
closely free FMN in water, the differences being < 0.5 ppm. Therefore it is
proposed that the polar part of the isoalloxazine (i.e. the pyrimidine sub-
nucleus including the N(5) atom) is fairly exposed to solvent. It should be
mentioned that hydrogen bonds alone are not sufficient to polarize the iso
alloxazine ring in such an extent. A relatively high dielectric constant is
needed additionally. The relative exposure to solvent would probably lead to
an increased mobility of the isoalloxazine ring, which is in agreement with
time-resolved fluorescence depolarization measurements (Visser et al_., 1983).
The exposure to solvent is also in agreement with the study of Claiborne et al.
(1982).
Upon binding of substrate pHBH (more than 90% of ES is present under
the conditions of these experiments) a drastic upfield shift of 2.9 ppm for
C(2) is observed. The observed value of 156.6 ppm is even close to the value
for the corresponding atom in flavin dissolved in chloroform (i.e. 155.2 ppm).
This shows that the TT electron density at C(2) is increased in the presence
of substrate, which indicates that the carbonyl group at C(2) is hardly po
larized. The resonance position of C(4) indicates that in the ES complex
a fairly strong hydrogen bond exists towards 0(4a). The upfield shift of the
line due to C(10a) upon substrate binding might be caused by the increased
TT electron density at C(2) (mesomeric effects). The upfield shift of the
resonance due to C(4a) suggests that the TT electron density is increased
at this position in the ES complex. It is, however, necessary to support this
15 by N NMR data. Nevertheless, the changes upon substrate binding clearly
show that the interaction between the prosthetic group and the apoprotein
is affected by the binding of the substrate. This is in agreement with changes
observed in adsorption and fluorescence spectra (Husain and Massey, 1979). Crys
tal lographic data of pHBH are at the moment only available for the enzyme-substrate
- 133 -
complex (Wierenga et al_., 1979). From crystallographic data (resolution 0.25 nm)
it has been suggested that hydrogen bonds exist to the N(l), 0(2a), N(3) and
0(4a) atoms of flavin. It should be mentioned that crystallographic data do not
allow to determine the strength of the hydrogen bonds, but allow only a rough
estimation of the distance between the atoms involved in hydrogen bond formation.
In contradiction to the crystallographic results the hydrogen bond to 0(2a)
13 must be weak as concluded from C NMR data. From crystallographic data it is
shown that an a-helix points with its positive pole directly to N(l) in the
ES complex (Wierenga et a]_. > 1979). This could be the reason why the C(2)
atom accomodates a rather high TT electron density. If this is true then the
relative position of the a-helix and the isoalloxazine ring is different
between the free enzyme and its ES complex. Mechanistic implications will
be discussed below.
It!§_r§dyç§<Lstate
In the absence of substrate the chemical shift of C(2) and C(10a) (Table 3)
clearly show that reduced FAD is present in the anionic form (FADH-). Com
pared with FMNH", C(2) and C(4a) are upfield shifted by 1.3 and 1.5 ppm respec
tively, wheras C(4)and C(10a) resonate at the same frequency as in FMNH". The
upfield shift of C(2) indicates that no or only a weak hydrogen bond exists
towards 0(2a). The upfield shift of C(4a) is probably caused by a strong TT 2
electron donation from N(10), i.e. N(10) is sp hybridized. The flavin in
oxidized free enzyme closely resembles free FMN in water, that in reduced
free enzyme does not resemble free FMNH". For the oxidized system it is
proposed that the C(2) of flavin gains IT electron density upon substrate binding,
due to some displacement of the a-helix. If this is true, then it is also
probable that the upfield shift of the resonance of C(2) in reduced free enzyme
is also caused by the position of the a-helix, i.e. pointing directly to
wards N(l). This implies that the binding mode of FADH" in free enzyme is
similar to the binding mode of oxidized FAD in the ES complex, but different
in the free oxidized enzyme.
134 -
•
A
B
A
D
180 160 140 120 100
PPM (6)
13 Fig. 4. Low field part of the C NMR spectra of p-hydroxybenzoate hydroxylase in which natural FAD was replaced by FAD, isotopically (>90%) enriched with 1 3C at the position C(2) and C(4a). A, oxidized state, B, oxidized state in the presence of 5 mM substrate, C, reduced state, D, reduced state in the presence of 5 mM substrate. Resonances marked with A represent natural abundance intensity of substrate. Resonances marked with a and • represent the C(2) and C(4a) atom, respectively. See caption of Fig. 3 for further explanations.
- 135 -
Table 3. 1 3 C chemical shifts of the carbon atoms C(2), C(4), C(4a) and C(10a)
of FAD bound to parahydroxybenzoate hydroxylase in the oxidized (ox)
or fully reduced (red) state and in the absence (-).or presence (+)
of substrate. Precision of the data is ± 0.2 ppm.
Redox state
ox
ox
red
red
Substrate
-
+
-
+
Carbon
2
159.5
156.6
156.9
156.9
atom
4
163.2
162.8
157.6
158.1
4a
136.4
135.7
99.9
99.3
10a
151.6
151.1
155.6
153.3
Upon binding of the substrate to the reduced enzyme the C(2) resonance
position is unchanged, which supports the above mentioned proposal, i.e. the
a-helix causes the upfield shift of C(2) and its relative position to the iso-
alloxazine ring is unchanged upon binding of the substrate to the reduced
enzyme. Upon binding of substrate to the reduced enzyme a remarkable upfield
shift of C(10a) is observed (2.3 ppm). According to the theory of Moonen et al_.
(1984) it is most likely that the upfield of C(10a) is caused by the N(5) atom of
flavin which is probably forced into the molecular plane, becoming a strong
•n electron donor. It should be noted that N(10) is already in the molecular
plane in the presence and absence of substrate. As deduced from the chemi
cal shift of C(10a), the hybridization of N(5) in the absence of substrate is
similar to that in FMNH" (i.e. the endocyclic angle of N(5) is about 115-117°).
Upon substrate binding the N(5) likely attains full sp2 hybridization. For C(4a)
a slight increase in -n electron density is observed upon substrate binding to
the reduced enzyme.
- 136 -
9.4.4. Mechanistic Implications
9.4.4.1. Substrate binding
From crystallographic data (Wierenga et al_., 1979) it is known that the
substrate is bound in the interior of the enzyme, very close to the isoalloxa-
zine ring. If the enzyme structure in the ES complex would be maintained in
the free enzyme the formation of the ES complex would be expected to be in
hibited due to steric effects. The results suggest that the structure of the ES
complex is not maintained in the free enzyme. In contrast, the data suggest a
rather dynamic active center, in which the isoalloxazine ring is exposed to solvent.
It is proposed that this feature benefits the complex formation between enzyme
and substrate.
Contrary to the free oxidized enzyme, the free reduced enzyme likely posesses
a structure similar with that of the oxidized and reduced ES complex. If
the proposal with respect to the free oxidized enzyme is correct (facili
tation of binding of substrate) then it would be expected that binding of
substrate to reduced free enzyme is much slower, as the active center con
formation is already pre-formed in reduced free enzyme. Indeed, the rate
of formation of the ES complex has been determined to be slower by more than
10 000 fold for the reduced enzyme as compared to the oxidized one, whereas
the equilibrium constant (K.) differs only 6-fold (Husain et aJL, 1980; Entsch
1 "i et al_., 1976). Thus, the combination of kinetic results and C NMR data strongly
suggest that the conformation of the active site is the same in the oxidized ES
complex, the free reduced enzyme and the reduced ES complex, but it differs in
conformation and mobility in free oxidized enzyme. The conformational change
probably also influences the observed dramatic increase of the reduction of FAD
by NADPH upon substrate binding, i.e. a high activation barrier for the redox
transition between the oxidized and the hydroquinone state might exist for free
enzyme, whereas the conformational change lowers this activation barrier.
- 137 -
9.4.4.2. Oxygen activation
Anionic reduced flavin has generally been regarded as a better oxygen
activator than the neutral from (Massey and Hemmerich, 1980; Favaudon, 1977).
Reduced pHBH is indeed anionic both in the absence and in the presence of sub
strate. The attack of dioxygen on the reduced isoalloxazine ring is a remarkable
reaction, as it violates the law of spin conservation. Nevertheless, reduced
flavins form rather smoothly a flavin hydroperoxide adduct both for free and some
protein-boun flavins. Among flavoproteins, however, the reaction rate differs dra
matically (cf. Chapter 1 and Massey et a]_, 1969). Oxygen reacts by an electrophilic
reaction. Thus, in principle, C(4a) and C(10a) could be regarded as the reaction
sites because the electron density is high at these centers (Moonen and Müller,
1984). Moreover these centers are less involved in resonance stabilization of
certain mesomeric structures of reduced flavin than e.g. the carbon atoms of the
benzene subnucleus. Indeed, in model systems both adducts are found. However,
in flavoenzymes, preferential attack on C(4a) occurs (Müller, 1983, and references
therein). As this reaction is an electrophilic reaction it is interesting to
compare the oxygen reactivity of various flavoproteins with the IT electron
13 density at the C(4a) center as measured by C NMR. In Table 4 the flavoproteins
investigated up till now are summarized.
The comparison should be regarded with caution, as for example steric effects
for the approach of oxygen are not taken into consideration. For example
flavodoxin from M.elsdenii flavodoxin is highly reactive towards oxygen
yielding 0„~, although the N(5) and C(4a) are shielded (chapter 11). The
C(4a) center in pHBH contains indeed a high u electron density as evidenced by
the high field position of the resonance due to C(4a). However, the IT electron
density at the C(4a) center is even higher in the 4 electron reduced lipoamide
dehydrogenase, but the reactivity with oxygen is low. On the other hand,
luciferase shows a 6(C(4a)) of 103.2 ppm indicating a decreased TT electron den
sity, although this protein reacts fast with oxygen and should in fact be regarded
as a hydroxylase. Thus from Table 4, no clear correlation between oxygen reacti-
- 138 -
Table 4: Comparison of reactivity of various reduced flavoproteins towards
oxygen versus the chemical shift of C(4a) in the reduced state. As
an example for flavodoxin, M.elsdenii flavodoxin is taken, for lipo-
amide dehydrogenase, the reader is referred to paragraph 9.3, for
luciferase from vibrio harveyi the chemical shift is reported by
Vervoort et ajh (1983), for a hydroxylase, pHBH from Pseudomonas
fluorescence is taken. The first product in the reaction of reduced
flavodoxin with 0« is 0~~. Therefore, for the other enzymes, as far as
appropriate, the reaction product in the absence of substrate is given.
Flavoprotein
flavodoxin
hydroxylase
lipoamide dehydrogenase
luciferase
6(C(4a))
ppm
103.5
99.3
98.9
103.2
Oxygen
reactivi
+
+
-
+
ty
Product
°2~ H 2 0 2
-
H 2 0 2
vity and IT electron density at the C(4a) center can be deduced.
It has been pointed out by some authors that electron transfer to
oxygen, yielding flavin radical and superoxide radical, precedes the formation
of a covalent adduct (Kemal and Bruice, 1979). If this is the case then indeed
addition to C(4a) would be expected by a radical recombination reaction, as
C(4a) possesses a high spin density in the semiquinone state (Müller, 1983).
It should be noted that the semiquinone (radical) state of FMN is planar
(Müller et al_., 1982). Therefore, it is anticipated that a planar hydroquinone
state has a higher reactivity towards oxygen than a bent one. If from a bent
flavin an intermediate semiquinone state is formed during the reaction with
oxygen a nuclear rearrangement must occur to accomodate the planar semiquinone
state. This nuclear rearrangement would probably cause an activation barrier
- 139 -
similarly to that as already discussed for the redox transitions in flavodoxin
(Moonen and Müller, 1982, 1984). In this respect it is interesting to note that
N(5) in reduced pHBH becomes more planar upon substrate binding. As also N(10)
is almost sp hybridized the reduced flavin in the ES complex is essentially
planar. From this it is postulated that the main feature of the reduced iso-
alloxazine ring to activate oxygen is not the IT electron density of the C(4a) 2
center, but rather the sp hybridization of the N(5) and N(10) atoms. The latter 2
atom is already essentially sp hybridized in FMNH , but the N(5) is only parti-2
ally sp hybridized in FMNH (^70%). As the hybridization of N(5) is hardly
influenced upon polarization of the flavin molecule, (Moonen and Müller, 1984)
full hybridization of the N(5) atom can probably only be achieved by steric
effects. It is interesting to note that it has been proven that in reduced
flavodoxin from M.elsdenii N(5) is forced into the molecular plane (Franken et
al., 1983). This flavodoxin shows indeed a high reactivity towards oxygen, yielding
superoxide anion, which is prevented by steric effects to form a covalent
adduct with protein-bound flavin. In luciferase the degree of hybridization
of N(5) is not known due to lack of sufficient NMR data. In lipoamide dehydrogenase,
13 it is estimated from the C NMR data that the N(5) atom is hybridized as in
FMNH", which could explain the lowered reactivity towards oxygen, according
to the postulate presented. The hybridization of N(5) could be of importance
for the problem of spin conservation in the reaction between reduced flavin and
triplet oxygen. The reaction would be spinail owed if not the singlet state,
but the triplet state of reduced flavin would be involved. It is however not
known if a low lying triplet state in reduced flavin exists. It would be of
great interest to solve this problem and to elucidate the influence of the va
rious conformations of reduced flavin on the triplet state.
- 140 -
REFERENCES
Bosma, H. (1984), Thesis Agricultural University, Wageningen, The Netherlands. Bruice, T.C. (1980), Acc.Chem.Res. 13, 256-262. Claiborne, A. and Massey, V. (1982) J.Biol.Chem. 258, 4919-4925. Cramer, F. and Neunhoeffer, H. (1962) Chem.Ber. 95T~1664-1669. De Kok, A., Visser, A.J.W.G. and De Graaf-Hess, TTC. (1982) in: Flavins and
Fl apoproteins (Massey, V. and Williams, C.H. Jr., Eds.), Elsevier North Holland Inc. New York, pp. 61-65.
Dixon, D.A., Lindler, D.L., Branchard, B. and Lipscomb, W.N. (1979) Biochemistry 18, 5770-5775.
Entsch, B., BalTöu, D.P. and Massey, V. (1976) J.Biol.Chem. 251, 2550-2563. Entsch, B., Husain, M., Ballou, D.P., Massey, V. and Walsh, ÏÏT~(1980)
J.Biol.Chem. 255, 1420-1429. Favaudon, V. (1777) Eur.J.Biochem 78, 293-307. Franken, H.D., Rüterjans, H., Müller, F. (1983) Eur.J.Biochem., in press. Husain, M. and Massey, V. (1979) J Bio!.Chem. 254, 6657-6666. Husain, M., Entsch, B., Ballou, D.P., Massey, V7~~and Chapman, P.J. (1980)
J.Biol.Chem. 255, 4189-4197. Kemal, C. and BruTce, T.C. (1979) J.Am.Chem.Soc. 101, 1635-1638. Ludwig, M.L, Burnett, R.M., Darling, G.D., Jordan7~"S~.R., Kendall, D.S.
and Smith, W.W. (1976) in: Flavins and Flavoproteins (Singer, T.P., Ed.) Elsevier, Amsterdam, pp. 393-404.
Massey, V., Gibson, Q.H. and Veeger, C. (1960) Biochem.J. 77, 341-Massey, V. and Veeger, C. (1961) Biochim.Biophys.Acta 48, 3~3- . Massey, V., Müller, F., Feldberg, R., Schuman, M., SulTTvan, P.A. Howell, L.G.
Mayhew, S.G., Matthews, R.G. and Foust, G.P. (1969) J.Biol.Chem. 244, 3999-4006.
Massey, V. and Hemmerich, P. (1980) Biochem.Soc.Trans. 8, 246-256. Matthews, R.G. and Williams, C.H., Jr., (1976) J.Biol.Cïïem. 251, 3956-3964. Matthews, R.G., Ballou, D.P., Williams, C.H., Jr. (1979) J.BTóT.Chem. 254,
4974-4981.
Mayhew, S.G. and Ludwig, M.L. (1975) in: The Enzymes (Boyer, P.D., Ed.) 3rd éd. Vol. 12, pp. 57-118, Academie Press, New York.
Mayhew, S.G. (1978) Eur.J.Biochem. 85, 535-547. Moonen, C.T.W, and Müller, F. (1982") Biochemistry 21, 408-414. Moonen, C.T.W, and Müller, F. (1984) Eur.J.Bioch"ê~m., submitted. Moonen, C.T.W., van den Berg, W.A.M., Boerjan, M. and Muller, F. (1984)
Biochemistry, submitted. Moonen, C.T.W., Vervoort, J. and Müller, F. (1984) Biochemistry, submitted. Müller, F., Voordouw, G., Van Berkel, W.J.H., Steennis, P.J., Visser, S. and
Van Rooijen, P.J. (1979) Eur.J.Biochem. 101, 235-244. Müller, F., Grande, H.J., Harding, L.J., Dunïïam, W.R., Visser, A.J.W.G.,
Reinders, J.H., Hemmerich, P. and Ehrenberg, A. (1981) Eur.J.Biochem. 116, 17-25.
Müller, F. and Van Berkel, W.J.H; (1982) Eur.J.Biochem. 128, 21-27. Müller, F. (1983) in: Topics in Current Chemistry 108, 7TrT07. Palmer, G. and Massey, V. (1968) in: BiologicaTUxTdlTtions (T.P. Singer, Ed.)
Wiley Interscience, New York, pp. 26"3~= . Searls, R.L., Peters, J.M; and Sanadi, D.R. (1961) J.Biol.Chem. 236, 2317-Shoun, H. and Beppu, T. (1982) J.Biol.Chem. 257, 3422-3428. Schulz, G.E., Schirmer, R.H., Sachsenheimer,~WT and Pai, E.G. (1978)
Nature 273, 120-Simondsen, R.P. and Tollin, G. (1980) Mol.Cell.Biochem. 33, 13-24. Smith, W.W., Burnett, R.M., Darling, G.D., Ludwig, M.L. TT977) J.Mol.Biol.
177, 195-225. Tauscher, L., Ghisla, S. and Hemmerich, P. (1973) Helv.Chim.Acta 56, 630-649. Van den Broek, H.W.J. (1971) Thesis Agricultural University, Wagenïngen
The Netherlands.
- 141 -
Van Leeuwen, J.W. (1983) Biochiin.Biophys.Acta, 743, 408-421. Van Schagen, C G . and Müller, F. (1981) Eur.J.Biochem. 120, 33-39. Vervoort, J., Ahmad, M., O'Kane, D.J., Müller, F. and Lee, J. (1983)
Biochem.Biophys.Res.Comm., in press. Visser, A.J.W.G., Penners, N.H.G. and Müller, F. (1983) in: Mobility and Recog
nition in Cell Biology (Sund, Veeger, Eds.) Walter de Gruyter & Co., Berlin] Watenpaugh, K.D., Sieker, L.C. and Jensen, L.H. (1976) in: Flavins and Flavo-
proteins (Singer, T.P., ed.) Elsevier, Amsterdam, pp. 405-410. Weyer, W.J. Hofsteenge, J. Vereijken, J.M., Jekel, P.A. and Beintema, J.J.
(1982) Biochim.Biophys.Acta 704, 385-388. Wierenga, R.K., De Jong, R.J.,Tälk, K.H., Hol, W.G.J. and Drenth, J. (1979)
J.Mol.Bio! 131, 55-73. Wilkinson, K.D. and Williams, C.H., Jr., (1979) J.Biol.Chem. 254, 852-862. Williams, C.H., Jr. (1976) in: The Enzymes (Boyer, P.D., Ed.) Vol. XIII,
Academic Press, New York, pp. 89-173. Wijnands, R.A. and Müller, F. (1982) Biochemistry 21, 6639-6646.
- 142 -
Chapter 10
A PROTON NUCLEAR MAGNETIC RESONANCE STUDY AT 500 MHZ ON MEGASPHAERA
ELSDENII FLAVODOXIN
A study on the stability, proton exchange and the assignment of some reso
nance lines.
Chrit T.W. Moonen and Franz Müller
SUMMARY
H NMR studies were performed on the three redox states of M.elsdenii
flavodoxin. The results show that the protein is remarkably stable, as con
cluded from amide proton exchange studies. Some amide protons are still present
in the lH NMR spectrum even after one month in 2H20 at 33°C (pH 8.3). The reac
tivity of the exchangeable protons can be grouped into three catagories, i.e.
t| » 5 min, 10s < t| < 5 min, and tj « 10s.
The amide proton exchange reactions are hardly dependent on the redox state. Op
timal resolution of H NMR spectra is obtained at 33°C, independent of the
redox state. No conformational change of the protein is observed in the pH range
between 6 and 8.5. Assignments of resonances to protons of flavin and of some
amino acid residues are established in both the oxidized and the hydroquinone state
using chemically and isotopically substituted flavins and the driven nuclear Over-
hauser technique. Preliminary two-dimensional H- H correlated spectra show
that the protein is amenable to two-dimensional NMR techniques. Previous assign
ments are confirmed by this technique.
- 143 -
INTRODUCTION
The flavodoxins form a class of small proteins (M 15.000-23.000) con
taining flavin mononucleotide (FMN) as the prosthetic group and serve as re
dox carriers in the living cell [1]. During in vitro redox reactions the pros
thetic group shuttles between the quinone (oxidized), semiquinone (radical)
and hydroquinone (1,5-dihydro or fully reduced) states. The redox potential
of the different redox states is strongly altered as compared with free FMN.
This can probably be ascribed to the influence of charges present in the active
site (Moonen and Müller, unpublished results). The crystallographic structure of
flavodoxin from Clostridium MP in its three redox states is known [2-4]. The
flavodoxin from M.elsdenii is strongly related to the former flavodoxin in many
chemical and physical aspects [1].
M.elsdenii flavodoxin, due to its easy availability in large quantities, has
been subjected to various NMR techniques. It has been shown that specific hy
drogen bonds are formed on binding of the prosthetic group to the apoprotein
[5,6]. The isoalloxazine ring (i.e. the redox-active moiety) of FMN appears to have
negligible motional freedom in the complex [7]. Moreover, it has been shown,
that a tryptophan residue (Trp-91) is located near the prosthetic group in the
oxidized complex. Van Schagen et a^. [8] and Moonen et aj[. [9] assigned all reso
nances of this tryptophan residue and showed that the internal flavin is shielded
from solvent by this tryptophan. The NMR studies also showed that the active
centers of M.elsdenii and Clostridium MP flavodoxins are identical with respect
- 144 -
to the structural arrangement of the flavin and the tryptophan residue [9].
A long recognized fact from the interchangeability with ferredoxins is that
only the transition between the semiquinone and the hydroquinone state in
Clostridium MP and M.elsdenii flavodoxin is of biological relevance [1]. Moonen
and Müller [10] showed that this is related to the fact that the activation
energy for the transition between the oxidized and the semiquinone state is
larger than that between the semiquinone and the hydroquinone state. According
to the crystallographic studies of Ludwig et al_. [4] on Clostridium MP flavo
doxin the different activation barriers can be rationalized by a minor confor
mational change in the transition from the oxidized to the semiquinone state.
Although the crystallographic data of Clostridium MP flavodoxin have pro
vided detailed information on the structure, one has to be careful with regard
to the dynamics of the given structures, especially since the flavin is located
near the surface of the protein. In addition protein-protein contacts in crystals
can influence the structure of proteins, as has been shown for the basic pancreatic
tryspin inhibitor by H NMR [11]. The recent enormous advances in two-dimensio
nal NMR techniques have even increased the potentialities of H NMR [12-15]. At
present the H NMR techniques even compete with crystallographic methods in
elucidating the structure of small proteins. The major advantage of the H NMR
methods may lie in the fact that they are in principle multi-parameter methods
by which structural and dynamic data can be acquired at the same time [16].
This prompted us to undertake a detailed H NMR study on M.elsdenii flavodoxin.
In this paper we describe the exchangeability of amide protons against deuterons
yielding information about the structural stability of the protein.
In addition some advanced techniques were used to assign resonance lines due to
protons of groups in or close to the protein active center. Some priliminary re
sults were reported elsewhere [17].
- 145 -
MATERIALS AND METHODS
Flavodoxin from M.elsdenii was isolated and purified according to published
procedures [18]. Reduction and reoxidation experiments were conducted by the
addition of the desired amount of a dithionite solution to the anaerobic
solution of flavodoxin. Anaerobiosis was achieved by carefully flushing the
solutions in the NMR tube with argon for about 20 min. The desired degree of
reoxidation was obtained by injecting small volumes of air into the NMR tube
containing the anaerobic reduced solution followed by gentle mixing. Apo-
flavodoxin was prepared as described by Wassink and Mayhew [19]. The reconsti
tution was performed at pH 7 by addition of an excess of FMN followed by ex
haustive dialysis. Wilmad 5 mm tubes of highest precision were used for all H
15 NMR experiments. Wilmad 10 mm precision tubes were used for the N NMR measure-
15 ments. The N NMR measurements and the experiments using the Redfield pulse
sequence were performed on a Bruker CXP-300 spectrometer. DOUBTFUL experiments
were done on a Bruker HX 360 spectrometer. All other spectra were recorded on
a Bruker WM 500 spectrometer. Normally 8K data points were used unless otherwise
stated. H Chemical shifts are reported relative to trimethylsilylpropionate
(TSP).
Samples contained 3 mM flavodoxin in 100 mM potassium phosphate buffer
(pH 8.3), except for the samples used for the two-dimensional technique and the
15
N NMR measurements which contained 8 mM flavodoxin in the same buffer. Measure
ments were performed at 33°C, except for the temperature dependent studies. To
reduce the water line in the NMR spectra, 100% H-0 was used as a solvent in most
experiments. When observation of labile protons was desired, a mixture of 10% 2
H20 and 90% H,,0 was used. In this mixed solvent the water line was diminished using
two different techniques depending on the kind of experiment: i) continuous irra
diation at the frequency of the water line except during the acquisition time,
ii) by the Redfield pulse sequence [20] resulting in a 'hole' in the excitation
spectrum at the frequency of the water line. The latter technique was slightly
- 146 -
modified for our experiments in that the normal sequence 2(+x)l(-x)4(+x)l(-x)
2(+x) was improved by optimizing both the length and the phase of the -x
pulse. This slight modification resulted in a drastically decreased exci
tation of the water hydrogens. A further advantage of the method is the dis
appearance of phase distortions in the resulting spectra.
Amide proton exchange was measured by following the decrease of the in-
tensity of the amide protons after dissolving the protein in H^O. At least
eight spectra were used for the determination of the rate constant of the
exchange reation.
Driven NOE difference spectra were recorded using the method of Dubs et
aJL [21]. An attenuation of 33 dB from 0.2 Watt was used for the saturation of
a particular resonance line.
DEPT spectra were recorded using the method of Doddrell et aj_. [221> °P~
timized for the J(NH) coupling constants reported by Franken et al_. [6]. The
selective decoupling power after the DEPT pulse sequence was 60 dB attenuated
15 from 20 Watt. The selective enrichment of FMN by the N isotope at the positions
N(l), N(3) and N(5) was achieved as described previously [23]. DOUBTFUL measure
ments were performed as described by Höre et al_. [24]. The two dimensional COSY
spectrum was obtained using the pulse sequence described by Aue et al_. [25].
Groups of 16 recordings with different phases were added for each value of t,
as described by Nagayama et al_. [26]. The spectrum of Fig. 6 was obtained from
512 spectra of 2K data points with an increment in t, of 8Q ys. The carrier fre
quency was at the low field end of the spectrum. The solvent resonance irradi
ation (15 dB attenuation from 0.2 Watt) was applied at all times, except during
data acquisition. Before Fourier transformation of the time domain spectra, free
induction decays were multiplied by a sine function (both in the t-, and the t~ di
rection). Furthermore, the time domain spectra were expanded to 2048 points in the
tj direction and 4096 points in the t„ direction by zero-filling to end up with a
1024x 1024 data matrix in the relevant frequency domain. The spectrum is shown in the
- 147 -
absolute value presentation. The spectrum was further improved by a sym-
metrization procedure in which the values of two corresponding points in the
spectrum were replaced by the lowest value.
RESULTS AND DISCUSSION
The H NMR spectra of oxidized and reduced M.elsdenii flavodoxin in H„0
? and of the oxidized protein in HLO are shown in Fig. 1. Several resonance lines
which are all shifted from the random-coil position are observed at the extremes
of high and low field. These can therefore be regarded as conformational probes.
At first glance most of the resonance lines remain at the same position in both re
dox states. This indicates that no gross conformational change occurs upon two-
electron reduction of the protein, as already concluded from previous H NMR measure
ments [27,28]. A closer inspection of the spectra shows that some resonance
lines appear at a different field when going from the oxidized to the reduced
state of the protein. Considering that the ring current effect of the bound flavin
is undoubtedly different in the two redox states, one cannot conclude that the
change in resonance position of some lines is caused by a conformational change.
Such a conclusion requires a more detailed analysis of the NMR spectra. In
the semiquinone state several resonance lines broaden beyond detection due to
the paramagnetic center [10]. This observation will be used below to assign
resonance lines due to groups in the active center.
Although the spectra of Fig. 1 show a remarkable resolution for a protein
of this size (M = 15.000) it is evident that several resonance lines overlap.
This renders specific assignments rather difficult. It is therefore not surprising
that some previous assignments to flavin protons are incorrect [28], as will be
shown below. Due to strong overlap of some lines we endeavoured to acquire the
spectra at conditions optimized for the resolution, i.e. at the optimum tempe
rature and the highest available magnetic fields.
Temgerature_and_2H-degendent_studies
The temperature dependence of the H NMR spectrum of oxidized flavodoxin in H~0
148
^ ^ u W V ^ i u J ^
10 PPM
lia H NMR spectra at 500 MHz of 8 mM oxidized (Bï and reduced (A) M.elsdenii flavo-
doxin in 0.1 M phosphate buffer (pH 8.3) at 332C (solvent is hUOjT Spectrum C represents the oxidized protein in HJO (see text). All labile protons have been exchanged prior to the experiment. The arrow indicates the residual water line, which is omitted from the spectrum. The resolution of all spectra was enhanced by a Lorentzian to Gaussian transformation The small baseline distortion around 5 ppm is due to the water line, which is not ideally saturated.
- 149 -
was investigated between 8° and 40°C. Several resonance lines were followed through
out this temperature range. No change in chemical shift position could be detected
for most resonances due to aliphatic and aromatic protons clearly indicating that
there are no gross conformational changes in the temperature region investigated.
Almost all amide resonances shift upfield upon an increase in temperature, however.
The upfield shift varies with the different amide resonances, with a maximal
shift of < 0.15 ppm for a temperature increase from 8° to 40°C. This behaviour is
expected for amide resonances involved in hydrogen bonds [29] and does not indicate
conformational changes. In the aromatic part of the spectrum no gross changes could
be detected at all. In basic trypsin inhibitor, for example, temperature increase
has been shown to cause a simplification of some resonance patterns due to tyrosine
residues, i.e. from a AA'BB' to a simple AB pattern [30]. This phenomenon has
been explained as an increase of the rotational motion around the C(ß) -C(y) axis
of the tyrosine residue. The fact that we do not observe such changes for flavodoxin
indicates that this rotational motion is either slow or fast throughout the tem
perature region investigated. A characteristic, observed throughout the whole
spectra of the three redox states, is a change in apparent T ? relaxation as de
termined from the linewidth. Increasing the temperature from 8° to 33 C causes a
sharpening of the resonances, which is likely caused by a decrease of the rotational
correlation time of the protein. Above 33°C the resonance lines broaden with
increasing temperature suggesting that conformational averaging influences the
Tp relaxation rates. Consequently most experiments were performed at 33 C yielding
an optimal resolution.
Varying the pH value of the solution from 6 to 8.5 has no influence on the po
sition of individual resonances. It should be noticed that M.elsdenii flavo
doxin contains no histidine residues [31,32].
Amidegrotonexchange
The NH protons exchange against deuterons of the solvent dissolving flavodoxin
in H~0. The rate constants of this exchange reaction give an indication on the
- 150 -
intramolecular fluctuations in the protein, since a partial opening of the structure
of the protein is required for the exchange reaction to occur [33].
About 60% of the total amide protons present in M.elsdenii flavodoxin was
calculated to exchange slowly enough to be observed after 5 min of dissolution
of the protein. This number is close to values expected for globular stable pro
teins [29]. The number of 60% was derived by integration of the low field part
(> 7.5 ppm) of a spectrum recorded immediately after dissolution of the protein 2
in H2O. The integrated area of the NH peaks was subsequently normalized with the
integral of the well resolved resonance line due to the methyl group resonating at
-0.74 ppm.
In the oxidized as well as in the reduced state there are several NH groups
which show a remarkable activation barrier against exchange. Some NH protons are
still observed even after one month in H-0 and at 33°C (pH 8.3). Spectrum C in
Fig. 1 was in fact obtained after exchange of the labile protons in the apoprotein
and subsequent reconstitution of the protein. These facts suggest that M.elsdenii
flavodoxin is a remarkable stable protein.
Due to considerable overlap (Fig. 1) it was not possible to trace the individual
exchange reactions of all NH protons. However, dissolving the protein in H^O
leaves four well resolved NH proton resonances located at the extreme of the low
field part (> 10 ppm) in the NMR spectrum (data not shown). The corresponding
NH groups are hardly affected upon change of the redox state. The exchange rate
was carefully measured by integrating the peak area at different times, after
dissolution of the protein in H20. The results are summarized in Table 1 and
lead to the conclusion that the exchange rates for resonances 3 and 4 are similar
in the two redox states, whereas the rates are different for the resonances 1
and 2. In addition it cannot be excluded that the relative positions of peaks 1
and 2 are reversed in the two redox states.
It has been shown that all NH protons exchange rapidly in apoflavodoxin [34].
Of interest is that the association constant for apoflavodoxin and FMN is about
20 times higher for the quinone than for the hydroquinone state. This difference
in association constants is clearly not reflected in the NH exchange rates for
- 151 -
the two redox states. Apparently we are predominantly measuring the exchange rates
of the complexes rather than those of the apoflavoprotein. i ?
Recording the H NMR spectrum in a mixture of H20/H20 (1:9, by vol.) instead
of in pure hLO allows the detection of labile protons, which are already exchanged
Table 1. The rates of exchange of labile protons against deuterons of the solvent
in oxidized and reduced M.elsdenii flavodoxin at 7 C and p H = 8.3. The
exchange rate is expressed as the time needed to reduce the original inten
sity of the resonances by 50%. The accuracy is ± 20%. The ti values <10s 2
are extrapolated from saturation transfer experiments at 33°C and pH 8.3.
Chemical shifts of all resonances are reported for 33°C for the sake of
comparison with future experiments. Resonances indicated by an asterisk are
located within 1 nm from the isoalloxazine ring, as deduced from the spec
trum of the paramagnetic semi qui none state.
Resonance line
1
2
3
4
5
6
7
8
9
10
Oxidation
Oxidized
ppm (6)
11.41
11.31
10.85
10.33
10.55
10.21
10.00
10.33
10.16
--
state, chemical
h
3.0 h
10.8 h
13.1 h
29.7 h
< 5 min
< 5 min
< 5 min
< 10 s
< 10 s
—
shift and exchanç
Hydroquinone
ppm(ô)
11.50
11.40*
10.92
10.35
10.26*
10.02
9.97*
10.23
10.33
10.52
e rates
2
5.6 h
2.1 h
13.3 h
28.0 h
< 5 min
< 5 min
< 5 min
< 10 s
< 10 s
< 10 s
- 152 -?
within 5 min after dissolution of the protein in HLO. In this way three NH
protons, exchanging with a t1 < 5 min, could be studied. These protons to-2
gether with the slow ones, are seen in Fig. 1, where well resolved resonances
are present around 10 ppm (Table 1) which are absent in the spectrum recorded in 2
FLO (data not shown).
AA
HYDR0QUIN0NE
***r««r<*^^^ SEMiaUINONE
OXIDIZED
i—i—i—i—i—i—i—i—i—i—i—i—r "i i i r
11.8
Fig. 2.
11.4 11.0
1
10.6
PPM (6)
10.2
i—i—i—i—i—r
9.8
Low field part of the H NMR spectra of 3 mM M.elsdenii flavodoxin in the three indicated redox states in 0.1 M phosphate buffer (pH 8.3) at 33 C (solvent is H„0). The spectra were obtained by the modified Redfield pulse sequence (cf. Materials and Methods). Resonances marked with •, A and A represent labile protons exchanging with a tj of » Ihr < 5 min and <10s, respectively, (see also Table 1).
- 153 -
If labile protons exchange rather fast, i.e. with a lifetime even approaching
the T~ values, the resonances disappear from the spectrum by chemical exchange due
to saturation transfer from the irradiated water protons [16]. The excitation of the
water protons can be avoided by use of the Redfield pulse sequence [20] (cf. Ma
terials and Methods) allowing the observation of the fast exchanging labile protons
(ti<10s). This is demonstrated in Fig. 2 for the well resolved low field part of
the H NMR spectrum. Compared with Fig. 1, three additional lines are seen in Fig.
2. The results are summarized in Table 1. It cannot be excluded that some resonances
crossed in the spectrum upon reduction, which renders the connections difficult.
Reduction by one electron (i.e. formation of the paramagnetic semiquinone form)
broadens some resonances due to labile protons beyond detection (Fig. 2 ) . These
resonances are marked in Table 1 and are due to labile atoms localized close
(<lnm) to the flavin ring system.
Assignments
The analysis of the active center of M.elsdenii flavodoxin by H NMR spec
troscopy would be facilitated if the resonance frequency of the hydrogen atoms of
the bound FMN was known. The resonances due to the C(6)H and C(9)H atoms are sing
lets in the qui none as well as in the hydroquinone state which should somewhat
facilitate their assignments. As the prosthetic group can be easily removed from
the holoprotein and the reconstituted protein yields identical H NMR spectra, un
ambiguous assignments should be aided by reconstitution with modified flavins.
Thus on substitution of the natural prosthetic group by 7-methyl-10-ribityl-isoalloxa-
zine-5'-monophosphate (i.e. replacement of the methyl group at C(8) by a pro
ton) the singlet at 7.12 ppm disappeared from the spectrum and the difference
spectrum between spectra of native and reconstituted oxidized flavodoxin showed
two doublets at 7.12 and 7.37 ppm and a singlet at 6.75 ppm. These doublets
(lJ ~ 7Hz) likely originate from C(9)H and C(8)H of the flavin. With respect
to the free flavin the resonances of C(9)H and C(8)H are shifted upfield by 0.61
and 0.42 ppm, respectively. These upfield shifts are in agreement with crystal-
- 154 -
lographic data on the related flavodoxin from Clostridium MP [1] from which it
can be deduced that the ring current effect from Trp-90 is larger for C(9)H than
C(8)H. In addition the upfield shifts of C(9)H for flavodoxin containing the na
tural and modified prosthetic groups are very similar, supporting the assign
ments. The previous assignment of the line at 6.68 ppm to C(9)H is therefore
incorrect [28]. The difference spectrum also shows non-vanishing intensity for
all resonances of Trp-91 for the frequencies, see [9]) and of a few other resonances.
This effect is ascribed to a slight difference between the ring current effects of
the two isoalloxazine rings and shows that Trp-91 is indeed located very close
to the isoalloxazine ring, as demonstrated previously [9]. This effect renders
however, assignments in rather crowded regions difficult. The resonance line at
6.75 ppm is therefore tentatively assigned to C(6)H. The assignment of the 7a
and 8a methyl groups is, on the other hand, easy because in the region where these
resonances are expected [35] the H NMR spectrum of M.elsdenii flavodoxin is not
very crowded. In the region between 2.6 and 2.8 ppm two sharp singlets are ob
served. Significantly the singlet at 2.77 ppm is absent in the NMR spectrum of
flavodoxin containing the modified flavin. The resonances at 2.77 and 2.69 ppm
are therefore assigned to the 8a and 7a methyl groups of the isoalloxazine ring,
respectively (Table 2 ) . This also means that the previous assignment of the re
sonances at 2.36 and 2.32 ppm to these methyl groups must be incorrect.
To obtain NMR spectra of M.elsdenii flavodoxin in the hydroquinone state
care must be taken that no trace of semi qui none is present in the solution
because the electron exchange rate between the semi qui none and the hydroquinone
state can be very fast, as is the case under the conditions the spectra in this
paper were recorded. In the semiquinone state the hydrogen atoms of the isoalloxa
zine ring are expected to have linewidths above 1kHz due to the paramagnetic
center [10]. Therefore, the presence of only 1% semiquinone will increase the line-
width of the resonances of isoalloxazine hydrogen atoms with at least 10 Hz.
The reduction was performed with dithionite which redox potential decreases
- 155 -
with decreasing pH values [36]. It was therefore necessary to keep the samples
above or at a pH value of 8.0. To avoid the ambiguities arising from the diffe
rence in ring current effects between the natural and modified protein-bound
flavin we replaced the natural prosthetic group in M.elsdenii flavodoxin by
15 [1,3,5- N.J FMN with an isotopic enrichment of >90 atom %. Some assignments are
15 subsequently established by N NMR (Fig. 3 ) . It has been shown [6] that the exchange frequency of the protons at N(3) and N(5) of flavin is slow in reduced
1 15 1 flavodoxin as deduced from the observation of J ( N- H) coupling constants
15 Although N NMR is a rather insensitive method, due to the low gyromag-
15 netic ratio, the existence of a clear coupling of the N atom with the attached
covalently bound hydrogen atoms should enable polarization of these nitrogen
atoms using the DEPT method (cf. Materials and Methods). This approach appeared
to enhance the signal to noise ratio appreciably. The spectra of Fig. 3 were
obtained within a measuring time of one hour using the DEPT pulse sequence. The
frequencies of the covalently bound protons were identified using specific ir
radiation during the detection period after the DEPT pulse sequence. First high
power off resonance decoupling was applied to roughly identify the proton fre
quencies. Subsequently the irradiation frequency was varied in steps of 10 Hz
around the roughly determined frequency. An attenuation of 60 dB from 20W was
applied in the latter experiments. The spectra shown in Fig. 3 were obtained under
optimal decoupling conditions of the N(3) and N(5) atoms of reduced flavodoxin.
The N(3)H and N(5)H resonances are thus determined to be at 9.38 and 5.67 ppm,
respectively (Table 2 ) . In Fig. 3 a broad resonance line of small intensity is
also observed near 120 ppm, a frequency where peptide nitrogens resonate [37].
This broad line most likely represents the natural abundance resonance of peptide
groups. The N(l) atom of reduced protein-bound flavin carries no hydrogen atom and
15 can therefore not be observed in the N NMR spectra using the DEPT pulse sequence.
It should be noted that the DEPT pulse sequence was unsuccessful for the as
signment of the N(3)H group in oxidized flavodoxin in at pH 8.3 and 33°C, since
- 156 -
^ V y v M ^ ^ N A V ^ / 4 ^ « ^ M ^ ^
150 100
PPM (6)
50
Fig. 3 Tî> —
N NMR spectra of reduced M.elsdenii flavodoxin (8 mM) in 0.1 M potassium phosphate buffer (pH 8.3) at 33°C (solvent is H 9 0). Spectra were obtained with the DEPT pulse sequence (cf. Materials and Methods). FMN was enriched (>902S) with ,th
N isotope at the positions N(l), N(3) and N(5). In the upper spectrum the 1JVLb
1 H ) was specifically decoupled and in the lower spectrum that of 1J(15N(5)-1H). 1 5 N chemical shifts are reported relative to NH,. For further details see text.
he N(3)-The
157 -
15, the N NMR line due to the N(3) atom appeared as a broad singlet owing to a
high exchange rate of the proton at N(3) with solvent water [6].
• < v 4 ^ ^
7.6 7.0 6À 5.8
PPM («]
Fig- 4 , The aromatic part of H NMR spectra of the hydroquinone state of 3 mM M.elsdenii flavodoxin in 0.1 M potassium phosphate buffer (p H 8.3) at 33 C (solvent is
HUO) is shown in the upper spectrum. The lower spectrum represents a difference spectrum between the upper spectrum and a similar spectrum from a solution consisting of about 2% semiquinone and of 98% of hydroquinone. For further details see text.
- 158 -
In an attempt to assign resonance lines to other protons of the reduced pro
tein-bound flavin we made use of the very fast electron exchange reaction be
tween the semiquinone and the hydroquinone state [10] and recorded difference
NMR spectra between flavodoxin in the reduced form and mixtures of reduced flavo-
doxin containing a few percent of semiquinone. Such a difference spectrum is shown
in Fig. 4 for only the low field part of the spectra. Only a few resonances remain
in the difference spectrum. The protons due to these resonances must be located
within about 0.8 nm from the paramagnetic flavin. Two sharp singlets remain in the
difference spectrum at 2.10 and 2.09 ppm (data not shown). These resonances are ten
tatively assigned to the methyl groups at position 8a and 7a, respectively, of the
prosthetic group. The resonances due to C(6)H and C(9)H are singlets and are expected
to appear in the region between 6 and 7 ppm. In Fig. 4B singlets are observed at
7.09, 6.50, 6.39 and 6.09 ppm.
Further discrimination between the four resonances can be made by the de
termination of interproton resonances by measuring the cross-relaxation rate,
manifested by the nuclear Overhauser effect (N0E). Although steady state H- H N0E
effects in proteins are rather unspecific due to spin diffusion [21], Kalk and
Berendsen [38] have shown that the initial build up of the N0E is inversely
proportional to the sixth power of the distance between the irradiated and the
observed hydrogen atom. 'Driven N0E' experiments [39] on oxidized M.elsdenii
flavodoxin are shown in Fig. 5. The resonance at -0.74 ppm is irradiated in these
spectra. This line has already been assigned to one of the 6-methyl groups of Leu-62
[28,10]. Specific NOE effects are observed in the difference spectra. Applying
irradiation times >ls renders the effects less specific. The closest hydrogen
atoms appear at -0.23 ppm (the other 5-methyl of Leu-62 [28]), at 1.10, 4.01 and
6.32 ppm. Upon performing the same experiments for the hydroquinone state the
resulting difference spectra show an NOE pattern extremely similar to that of
the oxidized state, although the resonance positions are slightly changed, i.e.
at -0.23, 1.06, 4.17 and 6.39 ppm. This observation also indicates that the
159 1, Table 2. Assignments of isoalloxazine and some other hydrogen atoms in the H
NMR spectra of oxidized and reduced M.elsdenii flavodoxin. Chemical shifts marked
with an asterisk are tentative assignments.
Proton(s)
FMN
N(3)H
N(5)H
C(6)H
C(9)H
CH3(8a)
CH3(7a)
Others
C(2)H, Trp-
C(2)H, Trp-
5-CH3, Leu-
6-CH3, Leu-
Y-CH , Leu-
a-CH , ?
C(4)H, Trp-
C(5)H, Trp-
C(6)H, Trp-
C(7)H, Trp-
-96
•91
-62
•62
•62
•91
91
91
91
Chemical
Oxidized
?
-
6.75*
7.12
2.77
2.69
6.32
-
-0.74
-0.23
1.10
4.01
7.61
6.80
5.57
6.41
shifts (ppm) in M.elsdenii flavodoxin
Reduced
9.38 ± 0.05
5.67 ± 0.05
6.50*
6.09*
il«
2.10
2.09*
6.39
7.09*
-0.23
1.06
4.17
—
—
—
—
- 160 -
conformation of the protein represented by the groups of these resonances
is similar in both redox states. It should be noted that the small difference
in frequency for the corresponding hydrogen atoms in the two redox states can
be explained by an attenuated ring current effect of the flavin upon reduction
[28]. In a previous paper [10] the distance between the 6-methyl group of Leu-62
J I I L 5 4
PPM (6I
Fig- 5 , Driven nuclear Overhauser H NMR spectra of 3 mM M.elsdenii flavodpxin in the Qxidized state in 0.1 M potassium phosphate buffer (p^H 8.3) at 33 C (solvent is
H 2 0). Upper spectrum represents a conventional ^H NMR spectrum. The lower spectra are difference spectra resulting from the subtraction of a 1 H NMR spectrum in which the high1field resonance was irradiated for a time T indicated in the Figure and a similar 1H NMR spectrum in which the irradiation frequency was shifted 200 Hz upfield from the first spectrum. Note that in the original spectrum in 2 H ? 0 several amide protons are present although the sample was kept 24 hours at 336c (p^H 8.3).
(-0.74 ppm) and the paramagnetic flavin was determined to be 1.02 nm. As a direct
N0E effect can only be expected for hydrogen atoms at distances of < 0.3 nm
from the irradiated group it can be concluded that the singlet at 6.39 ppm in
the hydroquinone and the corresponding singlet at 6.32 ppm in the oxidized state
cannot represent the C(6)H or the C(9)H atoms of flavin. A possible candidate
for the resonance at about 6.3 ppm would be the C(2)H atom of Trp-96. This as-
- 161 -
signment is in accordance with the three-dimensional structure of the related fla-
vodoxin from Clostridium MP [2-4]. The resonance is, as expected, also present in the
difference spectrum of Fig. 4. The correctness of the assignment is further supported
by the fact that the resonances belonging to Trp-91 have recently been assigned
[9]. None of these resonances appear around 6.3 ppm. Based on similar arguments
the lines at 1.1 ppm and at 4.01 and 4.17 ppm in the spectrum of oxidized and reduced
flavodoxin respectively, are assigned to y-CH of Leu-62 and a yet unidentified
a-CH group of an amino acid residue respectively. These assignments reduce the
possible candidates for the three residual singlets at 7.09, 6.50 and 6.09 ppm in
reduced flavodoxin (Fig. 4 ) . The resonance at 7.09 ppm most probably belongs to
C(2)H of Trp-91, in line with previous assignments [9]. The remaining two reso
nances are tentatively assigned to C(9)H and C(6)H of the prosthetic group
(Table 2).
Apart from the knowledge of interproton distances, the identification
of the resonances due to the different amino acids must be known for a successful
study of the structure [12]. As the spin system of an amino acid side chain
is rather specific for each amino acid, the identification can be done by the
determination of the through-bond 'J' connectivities throughout the side chain.
Van Schagen and Müller [28] have shown that specific irradiation (decoupling)
can be succussfully applied to achieve this goal. By this technique it has been
shown [28] that the y-CH group of Leu-62 resonates at about 1.1 ppm. However,
since that time powerful two-dimensional NMR methods have been developed to ar
rive at a map of through-bond connectivities of the whole H NMR spectrum in one
two-dimensional spectrum. In Fig. 6 an example of a so-called COSY spectrum (cf.
Materials and Methods) is shown of oxidized M.elsdenii flavodoxin in H20 to de
monstrate the applicability of this method. All off-diagonal peaks ('cross peaks')
show J-coupling between the resonances obtained by horizontal and vertical pro
jection of the cross peaks onto the diagonal, which represents the normal one-
-dimensional NMR spectrum. The clear pattern observed in Fig. 6 is encouraging
162
JWA^
Q
-3
°Â
-tm—
O
D
3 A
0
0
o
e
ß ©,-
8r V ^ ^ ^
(o®
i
2S^
o----
l D
Ar i * r
! ^ 1 0 1
-o
9-i i i i
i
1 Ji
i
fa ' -?
Q '
Q
o °
s o
0
• r
- to
C--
oo
8 7 PPM (s)
Fig. 6 1, Aromatic part of the contour plot of the COSY H NMR spectrum of oxidized M.elsdenii flavodoxin in 0.1 M potassium phosphate buffer (pH 8.3) at 33 C (solvent is HpO) (cf. Materials and Methods). For comparison the one-dimensional NMR spectra are also shown at the upper and left part of the figure. These one-dimensional spectra are optimally resolution-enhanced by a Lorentzian to Gaussian transformation. For experimental details see Materials and Methods. The outlined connectivity pattern represents the benzene part of the indole ring of Trp-91 (for the assignment, see [9]. The diagonal represents the conventional one-dimensional spectrum.
- 163 -
to start a detailed NMR study on this protein. This result was not obvious,
beforehand, as line widths are larger than for the proteins investigated with
these methods so far. During the transfer of magnetization by J-coupling, loss
of magnetization due to T ? relaxation is inevitable. Therefore, on increasing
the size of the protein (and thus the T 2 relaxation times) the loss of magnetization
during the COSY pulse sequence is also increased. Nevertheless, with this flavodxin
it appears to be possible to arrive at strong cross peaks. In Fig. 6 the low
field part of a COSY spectrum is shown which is chosen to compare it with results
obtained earlier by a different technique [9], The feature seen in Fig. 6 is the
coupling pattern of the benzene part of the indole moiety of Trp-91 (Table 2 ) ,
which is in perfect agreement with results obtained by a time-resolved photo-
chemically induced dynamic nuclear polarization (CIDNP) study [9]. Trp-91 was shown
to be at the surface of the protein but interacting strongly with the bound fla
vin, thereby shielding the isoalloxazine ring from solvent water. This ex
plains the unusual shifts of the hydrogen atoms of the indole ring of Trp-91.
Furthermore, the connectivity between the resonances near 6.66 and 7.06 ppm
likely belong either to one or two tyrosine residues . This connectivity is con
firmed by the DOUBTFUL technique (data not shown), which is based on double
quantum coherence. In addition this tyrosine residue is not polarized by the CIDNP
technique. Whether the C(3)H and C(5)H atoms of this tyrosine residue are
magnetically slightly"inequival ent, due to a limited rotation around the
C. - 6 axis, or whether two tyrosine residues are overlapping could not be
determined unambiguously.
The 'Driven NOE' experiments and the preliminary COSY experiment show that
this flavodoxin is amenable to a detailed two-dimensional H NMR study which
will be reported later. We hope to be able to solve the structure of the active
cluster of M.elsdenii flavodoxin in detail by these techniques and to report the
results in the near future.
- 164 -
REFERENCES
1. Mayhew, S.G. and Ludwig, M.L. (1975) Enzymes, 3rd Ed. 12, pp. 57-118. 2. Burnett, R.M., Darling, G.D., Kendall, D.S., LeQuesne, M.E., Mayhew,
S.G., Smith, W.W. and Ludwig, M.L. (1974) J.Biol.Chem. 249, 4383-4392. 3. Smith, W.W., Burnett, R.M., Darling, G.D. and Ludwig, M X 7 (1977)
J.MolI.Biol. 117, 195-225. 4. Ludwig, M.L., Burnett, R.M., Darling, G.D., Jordan, S.R., Kendall, D.S.
and Smith, W.W. (1976) in: Flavins and Flavoproteins (Singer, T.P. ed.) Elsevier Scientific Publishing Co., Amsterdam, pp. 393-403.
5. Van Schagen, C G . and Müller, F. (1981) Eur.J.Biochem. 120, 33-39. 6. Franken, H.D., Rüterjans, H. and Müller, F. (1983) Eur.JTB~iochem.,
in press. 7. Moonen, C.T.W, and Müller, F. (1983) Eur.J.Biochem. 133, 463-470. 8. Van Schagen, C G . , Müller, F. and Kaptein, R. (1982)~B~Tochemistry 21,
402-407. ~~' 9. Moonen, C.T.W., Höre, P.J., Müller, F., Kaptein, R. and Mayhew, S.G.
(1982) FEBS Lett. 149, 141-146. 10. Moonen, C.T.W. and~Mïïller, F. (1983) Eur.J.Biochem., submitted. 11. Brown, L.R., De Marco, A., Richarz, R., Wagner, G. and Wüthrich, K.
(1978). Eur.J.Biochem. 88, 87-95. 12. Wüthrich, K., Wider, G., Wagner, G. and Braun, W. (1982)
J.Mol.Biol. 155, 311-319. 13. Bill eter, M.~TB>aun, W. and Wüthrich, K. (1982) J.Mol .Biol. 155
321-346. 14. Wagner, G. and Wüthrich, K. (1982) J.Mol.Biol. 155, 347-366. 15. Wider, G., Lee, K.H. and Wüthrich, K. (1982) J.MÔT. Biol. 155, 367-388. 16. Jardetzky, 0. and Roberts, G.C.K. (1981) in: NMR in Molecular Biology
Academic Press, Ney York, pp. 11-68. 17. Müller, F. and Moonen, C.T.W. (1982) in Flavins and Flavoproteins
(Massey, V. and Williams, C.H., Eds.) Elsevier North Holland, New York, pp. 517-527.
18. Mayhew, S.G. and Massey, V. (1969) J.Biol.Chem. 244, 794-802. 19. Wassink, J.H. and Mayhew S.G. (1975) Anal.Biochem. 68, 609-616. 20. Redfield, A.G., Kunz, S.D. and Ralph, E.K. (1975) JTTfogn.Reson. 19,
114-117. 21. Dubs, A., Wagner, G. and Wüthrich, K. (1979) Biochim.Biophys.Acta
577, 177-194. 22. EöcTdrell, D.M., Pegg, D.T. and Bendall, M.R. (1982) J.Magn.Reson. 48
323-327. 23. Müller, F., Vervoort, J., Lee, J., Horowitz, M. and Carreira, L.A.
(1983) J.Raman Spectrosc. 14, 106-117. 24. Höre, P.J., Zuiderweg, E.RTF., Nicolay, K., Dijkstra, K. and Kaptein,
R. (1982) J.Am.Chem.Soc. 104, 4286-4288. 25. Aue, W.P., Bartholdi, E. and" Ernst, R.R. (1976) J.Chem.Phys. 64,
2229-2246. 26. Nagayama, K., Anil Kumar, Wüthrich, K. and Ernst, R.R. (1980)
J.Magn.Reson. 40 321-334. 27. James, T.L., Ludwig, M.L. and Cohn, M. (1973) Proc.Natl.Acad.Sci.USA
70, 3292-3295. 28. Vän Schagen, C G . and Müller, F. (1981) FEBS Lett. 136, 75-79. 29. Wagner, G. (1977). Ph.D. Thesis nr. 5992, ETH ZüricïïT" 30. Wagner, G., De Marco, A. and Wüthrich, K. (1976) Biophys.Struct.Mechanism.
2, 139-158. 31. Tanaka, M., Haniu, M., Yasunobu, K.T., Mayhew, S.G. and Massey, V. (1973)
J.Biol.Chem. 248, 4354-4366. 32. Tanaka, M., Haniu, M., Yasunobu, K.T., Mayhew, S.G. and Massey, V. (1974)
J.Biol.Chem. 249, 4397.
- 165 -
33. Wagner, G. and Wüthrich, K. (1978) Nature 275, 247-248. 34. Moonen, C.T.W. and Müller, F. (1982) Biochemistry 21, 408-414. 35. Grande, H.J., Van Schagen, C G . , Jarbandhan, T. ancTMüller, F. (1977)
Heiv.Chim.Acta 60, 348-366. 36. Mayhew, S.G. (1977) Eur.J.Biochem. 85, 535-547. 37. Levy, G.C. and Lichter, R.L. (1979)^Tn: Nitrogen-15 Nuclear Magnetic
Resonance Spectroscopy, John Wiley and Sons, New York, pp. 28-107. 38. Kalk, A. and Berendsen, H.J.C. (1976) J.Magn.Reson. 24, 343-366. 39. Richarz, R. and Wüthrich, K. (1978) J.Magn.Reson. 30, 147-150.
Acknowledgement
We are indebted to Mr. M.M. Bouwmans for the preparation of the figures,
Miss C M . Verstege for typing the manuscript, to Mr. W.A.M. van den Berg
for the isolation of the flavodoxin, Dr. C.A.H. Rasmussen for carefully reading
the manuscript, Dr. P. Hore for the help with the DOUBTFUL measurements and
Mr. A. de Jager for skilful technical assistance.
This study has been carried out under auspices of The Netherlands Foun
dation for Chemical Research (SON) with financial aid from the Netherlands
Organisation for the Advancement of Pure Research (ZWO).
- 166 -
Chapter 11
THE USE OF TWO-DIMENSIONAL NMR SPECTROSCOPY AND TWO-DIMENSIONAL DIFFERENCE
SPECTRA IN THE ELUCIDATION OF THE ACTIVE CENTER OF MEGASPHAERA ELSDENII
FLAVODOXIN
Chrit T.W. Moonen, Ruud M. Scheek, Rolf Boelens and Franz Müller
- 167 -
SUMMARY
^ H "through bond" correlated (COSY) and ^ H "through space" (NOESY)
two-dimensional NMR techniques were applied to study the structure of M.elsdenii
flavodoxin in the oxidized and reduced state. It is shown that two-dimensional
NOESY difference spectra between spectra of flavodoxin in the reduced and semi-
quinone state result in a spectrum representing only the active center of the
fully reduced state. The sphere of the active center observed in the difference
spectra can be varied easily by changing the relative amount of flavodoxin semi-
quinone in the second sample. The difference NOESY spectra simplified the analysis
of the complex spectra. Resonances could be assigned to Ala-56, Tyr-89 and Trp-91,
which are located in the direct vicinity of the protein-bound flavin. The relative
positions and side chain dihedral angles of these residues are compared for the
two redox states. Ala-56 and Tyr-89 show identical relative positions and dihedral
angels in the two redox states, although the rotational motion of Tyr-89 is en
hanced in the oxidized state. In both redox states Trp-91 is immobilized and ex
tremely close to the prosthetic group. However, a small displacement of Trp-91
towards the N(5) atom of the flavin occurs upon reduction. The results obtained
for Trp-91 are in excellent agreement with crystallographic results of the related
flavodoxin from Clostridium MP. However, the latter studies showed a somewhat dif
ferent position of the tyrosine residue compared with our results.
INTRODUCTION
Flavodoxin from Megasphaera elsdenii has a molecular weight of 15.000 daltons.
The amino acid sequence is known (Tanaka et al_., 1973, 1974). Flavodoxins contain
riboflavin-5'-monophosphate (FMN) as a prosthetic group and function as electron
carriers in biological reactions (Mayhew and Ludwig, 1975). The protein-bound
flavin can occur in the oxidized, the semiquinone and the hydroquinone state. In
biological reactions, however, flavodoxins do not make use of the oxidized state,
owing to a high activation barrier between the oxidized and the semiquinone state
(Moonen and Müller, 1982a, 1983a).
- 168 -
In a previous study we have shown by NMR methods that M.elsdenii flavodoxin
is remarkably stable in the three redox states over a long period of time (Moonen
and Müller, 1983b). It was concluded from these data that this flavodoxin appeared
to be amenable to two-dimensional NMR spectroscopy. Recent developments in two-
dimensional NMR spectroscopy (Wüthrich et a_K, 1982; Billeter et aj_., 1982; Wagner
and Wüthrich, 1982; Wider et al_., 1982) have shown that detailed structural infor
mation can be obtained for small proteins by these techniques (Steinmetz et al.,
1981). In this paper we report on a two-dimensional H NMR study with the aim
to elucidate the structure of the active center of M.elsdenii flavodoxin. Although
these two-dimensional NMR techniques have up to now only been applied to proteins
of a molecular weight less than 10.000 daltons our results show that these tech
niques can also be applied succesfully to larger proteins, in spite of the higher
complexity of the H NMR spectra. Since the protein-bound flavin forms a stable
radical we used this property of the protein in combination with difference NMR
spectra to obtain information with regard to the structure of the active center
of the protein.
MATERIALS AND METHODS
Flavodoxin from Megasphaera elsdenii was isolated and purified as previously
described (Mayhew and Massey, 1969). Wilmad 5 mm precision tubes were used. All
samples contained 8 mM flavodoxin in 0.1 M potassium phosphate, pH 8.3. A mixture 2
of 10% H20/90% H20 was used as solvent. Anaerobiosis was achieved by carefully
flushing the sample with argon for 20 min. The desired degree of reduction was
achieved by adding an appropriate volume of a dithionite solution (Mayhew, 1978).
Reduced or partially reduced samples were sealed after the reduction.
H- H correlated spectra were obtained by the "COSY" pulse sequence (90°-
t^-90 -t2) with an appropriate phase cycling to eliminate experimental artefacts
(Nagayama et al_., 1980).
H-^H N0ESY spectra were recorded with the sequence (90°-t1-90°-TM-90°-t2)
and the phase cycling as described by Anil Kumar et al_. (1980). Magnetization ex-
- 169 -
change occurs during the mixing time T „ for which normally 0.1 s was choosen,
unless otherwise stated.
All spectra were recorded on a Bruker WM 500 spectrometer operating at 500
MHz. Quadrature detection was used for data acquisition. The spectral width was
12500 Hz with the carrier frequency at the left extreme. Normally 512 spectra of
1024 points were accumulated. Specific irradiation of 13-17 dB attenuation from
0.2 Watt was applied to saturate the protons due to water. The water protons were
irradiated at all times, except during data acquisition. The repetition time needed
to saturate the water protons sufficiently was 2.3 s. The acquisition of a two-
dimensional spectrum lasted 30 to 50 h. The temperature was 33°C.
After data acquisition, all further data handling was carried out on a CYBER
170/760 computer of the University of Groningen, Groningen, The Netherlands. Zero-
filling was applied to end up with a 512x512 point data matrix in the frequency
domain. Before Fourier transformation, the time domain data matrix was multiplied
with a sine bell for the COSY spectra and with a 45° shifted sine bell for the
N0ESY spectra, both in the t, and the t~ direction. The length of the sine bell
was adjusted so as to reach zero at the last experimental data point. After Fourier
transformation the absolute value spectrum was calculated for the COSY spectra,
whereas pure absorption spectra were obtained for the N0ESY spectra after phase
adjusting as described by States et al_. (1982).
For the two-dimensional difference spectra, a baseline correction was applied
first, omitting a region of about 0.3 ppm around the water line. Then the two
spectra were normalized on the basis of an "off-diagonal peak" which represents a
through space connectivity far from the active center. Subsequently, the subtrac
tion was performed.
RESULTS AND DISCUSSION
B§dyç§çLfl§y2<!2£in
A COSY spectrum of two-electron reduced M.elsdenii flavodoxin is shown in
Figure 1. The information in the region around 4.7±0.2 ppm is lost due to the
- 170 -
high power used to saturate the water protons. Several cross peaks are evident
in the NH.-aCH. region of the spectrum. In principle each amino acid residue
gives rise to one such peak, with the possible exception of proline and glycine,
and residues which are exposed to solvent. Some spin systems can be identified
from Figure 1. The doublet at 0.11 ppm for example has already been assigned to
the methyl group of Ala-56 (Moonen and Müller, 1983a). From the COSY spectrum we
find the a-CH resonance at 2.66 ppm and that of the amide proton at 6.47 ppm. The
doublets at -0.65 and -0.23 ppm have previously been assigned to the <5-CH, groups
of Leu-62 (Van Schagen and Müller, 1981). These methyl groups couple with the same
proton resonating at 1.06 ppm. The doublets (methyl groups) at 0.03 and -0.39 ppm
also couple with a common proton at 0.90 ppm. The triplet (one methyl group) at
-0.43 ppm probably represents the «5-CH, group of a lie residue as a coupling is
observed with two non-equivalent hydrogen atoms at 0.73 and 1.51 ppm. In the con
ventional H NMR spectrum (see Figure 1) three doublets and one triplet are ob
served between -0.15 and -0.45 ppm. Interestingly, the COSY spectrum clearly shows
that two additional resonances are present in this region, i.e. at -0.19 ppm,
coupled with a hydrogen atom resonating at 1.37 ppm, and a group at -0.36 ppm,
coupled with a hydrogen atom resonating at 1.20 ppm. These additional resonances
are "hidden" in the conventional NMR spectrum. Although several coupling patterns
can be demonstrated in Figure 1 unambiguous assignment of the complete spin sys
tem of a particular amino acid residue appeared to be possible only for simple
spin systems, e.g. the Ala-56. This is due to the fact that the region between
2.4 and 0.8 ppm is overcrowded, even in the two-dimensional spectrum (see below).
We expected to observe the coupling pattern of the indole moiety of Trp-91 in
the aromatic region of the spectrum (Figure 1 ) , as previously observed for the
oxidized protein by a different NMR technique (Moonen et aj_., 1982b). This pattern
is not found in Figure 1.
To aid further assignments the results of Figure 1 are complemented by re
sults obtained by through-space connectivities as observed by the determination of
cross-correlation rates by the nuclear Overhauser effect (Anil Kumar et al_., 1980).
The N0ESY spectrum (Figure 2 ) , obtained with a mixing time of 0.2 s, shows that
- 171 -
Vdj^yv^M
Figure 1: Absolute value "contour" plot of a H- H correlated (COSY) spectrum of the hydroquinone state of M.elsdenii flavodoxin in 0.1 M potassium phosphate buffer (pH 8.3). The solvent consists of 90% H„0 and 10% H20. The diagonal represents the normal one-dimensional spectrum. All off-diagonal "cross-peaks" represent a through-bond J-connectivity between the resonances found by horizontal and vertical projection onto the diagonal. For comparison the one-dimensional spectrum (resolution enhanced by a Lorentzian to Gaussian transformation) is shown in the upper part of the spectrum. For further details, see Materials and Methods.
172
10
• Oo • a ° . . ° <? %*° o » • ° O 9 o o » a o o » * • - O -
o ' . = " " . •" • , ° «o •
10 6 4 PPMI6)
Figure 2: Pure absorption (contour) plot of a H- H NOESY spectrum of the hydro-quinone state of M.elsdenii flavodoxin in 0.1 M potassium phosphate buffer (pH 8.3). Solvent consists of 90% H20 and,10| H-O. A mixing time T „ of 0.2s was used. Cross peaks represent H- H tnrough-space N0E effects between the resonances found by horizontal and vertical projection on the diagonal. For further details, see Materials and Methods.
the pure absorption representation results in a much better resolution than the
absolute value representation of the COSY spectra (Figure 1). Similar through
space contacts for well resolved resonances were also observed by conventional
driven N0E measurements (Moonen and Müller, 1983b) for some well resolved reson
ance lines and appear to be identical in the spectra obtained by both techniques.
In Figure 2 a large number of connectivities is observed, as expected for a pro-
173 -
tein of this size. For example, the S-CH, (0.11 ppm) of Ala-56 transfers its mag
netization to the a-CH (2.66 ppm) and the NH group (6.47 ppm) of the residue.
Sequential assignments are rather difficult however, again because of severe
overlapping of resonance lines. All the same a dramatic simplification of the
spectra can be achieved by utilizing the kinetic properties i.e. the negligible
activation energy of the electron exchange reaction between the semiquinone and
the hydroquinone states (Moonen and Müller, 1983a). In addition the structures
of the two redox states are very similar.
0
10
fc '£
10 6 4 PPM(Ö)
Figure 3: Pure absorption (contour) plot of a H- H NOESY spectrum of approximately 90% hydroquinone and 10% semiquinone state of M.elsdenii flavo-doxin in 0.1 M potassium phosphate buffer (pH 8.3). Solvent consists of 90% H20 and 10% H^O. A mixing time T,. of 0.1 s was applied. For further aetails, see Materials and Methods.
- 174 -
In flavosemiquinone most of the spin density is confined to the pyrazine
moiety of the isoalloxazine ring (Müller, 1983). Owing to the paramagnetism of
the one-electron reduced protein-bound prosthetic group the resonances due to
protons in the vicinity of the semi qui none will be broadened. Since the electron
exchange reaction is diffusion determined the presence of for instance 10% of the
semiquinone in the solution studied, has the same effect on the resonances as
placing 0.1 of an electron on the flavin of each protein molecule. In Figure 3
a NOESY spectrum is presented of a solution of M.elsdenii flavodoxin consisting
of about 90% hydroquinone and about 10% semiquinone. Despite the presence of the
paramagnetic center, it appeared to be possible to sufficiently saturate the water
protons, indicating that the flavin is fairly well shielded from the solvent. Most
of the connectivities observed in Figure 2 are also present in Figure 3. Other con
tacts are clearly removed from the spectrum. In fact, Figure 3 represents a NOESY
spectrum of the hydroquinone state minus the environment of the flavin in the
paramagnetic state. Two different mechanisms cause the disappearance of the active
center protons in Figure 3, i.e. broadening and a leakage mechanism occurs due to
the paramagnetism. The latter competes with the magnetization transfer for hydrogen
atoms near the flavin during the mixing time. This renders the effects of the para
magnetism more pronounced in the NOESY spectrum than in the COSY spectrum (data
not shown). In Figure 4 a difference two-dimensional NOESY spectrum is shown re
sulting from a subtraction of a spectrum of the hydroquinone state and of the spec
trum presented in Figure 3. Both experiments were performed under identical condi
tions (cf. Materials and Methods). Figure 4 is thus a NOESY spectrum of only the
active center of the hydroquinone state of M.elsdenii flavodoxin. It is estimated
that interproton connectivities in Figure 4 extend to a limit of about 1 nm from
the paramagnetic center. This distance can be adjusted by varying the concentration
of the semiquinone. In this way a series of difference spectra can be produced con
taining different structural information. At any rate it can be seen from Figure 4
that the problem of overcrowded regions is greatly reduced by this technique and
that assignments are facilitated. For example,the proton at N(5) of flavin resonates
at 5.67±0.05 ppm (Moonen and Muller, 1983b). A strong NOESY connectivity is found
175 -
6
10
» •
'<<•
, • / '
S
0
t. ...
* •: •» .
. . . . ' . / . ; •
- « / 'o
• -• •
• £ • * < . /y
\"/' >V' . • e fa* • ' . »l-ipf
• -X **. •
,
10 6 PPM15)
Figure 4: Difference H- H NOESY spectrum (contour plot) in pure absorption phase, representing the NOESY spectrum of only the active center of the hydroquinone state of M.elsdenii flavodoxin (sphere of the active center is estimated to be about 1 nm). The spectrum is obtained by a subtraction of a NOESY spectrum of 100% reduced and the NOESY spectrum presented in Figure 3. The two spectra were recorded under the same experimental conditions. Mixing time x was 0.1 s for both spectra. For further details, see Materials and Metnods.
between the resonance at 5.63 ppm and a singlet at 6.50 ppm. The proton at N(5)
of flavin is undoubtedly very close (<0.25 nm) to the proton at C(6) of flavin.
The singlet at 6.50 ppm is therefore assigned to C(6)H of the prosthetic group.
Using a mixing time of 0.2 s a connectivity is also observed between the resonance
at 5.63 ppm and that at 7.79 ppm, which is a doublet. This doublet is assigned to
- 176 -
C(2)H of a tryptophan residue, it shows a strong coupling (Figure 1) and a strong
NOE connectivity to a resonance at 6.94 ppm (a well resolved triplet). The triplet,
on the other hand, exhibits NOE connectivities with the line at 7.79 ppm and a
resonance line at 7.10 ppm. The latter is obscured by several overlapping resonances.
In Figure 1 the COSY connectivity between the resonances at 6.94 ppm and 7.10 ppm
is partially obscured by the broad diagonal (due to the absolute value presentation).
The obscured resonance line at 7.10 ppm shows clear COSY and NOE connectivities with
the well resolved doublet at 7.53 ppm. Thus the doublet at 7.79, the triplet at
6.94, the multiplet at 7.10 and the doublet at 7.53 ppm are assigned to Trp-91
(see Table 1). These assignments are further supported by the weak NOE connecti
vity between the resonances at 6.94 and 7.53 ppm, when a mixing time of 0.2 s is
applied (Figure 2 ) . Moreover from the presence of the NOE connectivities in
Figure 4 and previously published one dimensional difference spectra (Moonen and
Muller, 1983b) it must be concluded that the tryptophan residue is close to the
isoalloxazine ring system. Other resonances of Trp-91 can be assigned from Figures
2 and 4 . The group due to the resonance at 7.53 ppm exhibits a strong NOE con
nectivity with the resonance line at 3.50 ppm. Using a mixing time of 0.2 s a weak
NOE connectivity to the line at 3.28 ppm is observed. The two resonances at 3.50
and 3.28 ppm show a strong mutual NOE connectivity, also present in Figure 4, in
dicating a small distance to the flavin ring. In addition they show a NOE connec
tivity to a resonance at 5.04 ppm which in turn is close to the group resonating
at 7.53 ppm (Table 1). Thus, the resonances at 3.50 and 3.28 ppm represent the
two nonequivalent C(g) protons and the resonance at 5.04 ppm represents the C(a)
proton of Trp-91. The resonance line at 7.08 ppm is assigned to the C(2)H of
Trp-91 as the line at 3.28 ppm shows a strong NOE effect on that at 7.08 ppm. The
C(a)H group also exhibits a small NOE effect on the resonance at 7.08 ppm, though
only with a 0.2 s mixing time. It should be noticed that no COSY connectivity for
C(o)H-C(ß)H was observed (Figure 1). The absence of such connectivities is unfor
tunately a rather common feature for the technique used in Figure 1 (cf. below).
The intraresidue N0ESY connectivities allow a rough determination of the
dihedral angles Xi and x 2 of tne side chain of Trp-91, like previously calculated
- 177 -
from time-resolved photochemically induced dynamic nuclear polarization spectra
(Moonen et al_., 1982). The results agree most with Xl = 60°C and x2 = ± 9 0 ° c -
Further NOE connectivities are observed between the line at 6.59 ppm and the
resonances at 7.06 and 7.42 ppm (Figure 4 ) . The resonances at 6.59 and 7.42 ppm
are doublets, whereas the multiplet pattern of the resonance at 7.06 ppm could not
be determined due to severe overlapping. As both connectivities are also found in
the COSY spectrum (Figure 1) the resonances probably belong to a tyrosine residue
with non-equivalent C(2,6) protons and two overlapping C(3,5) protons. Another
strong NOE connectivity is found between the resonance line at 7.06 ppm and a
resonance at 3.12 ppm, and a weak one with the line at 2.63 ppm. The latter two
resonances show strong mutual NOE and COSY connectivities, indicating that they
represent the two non-equivalent C(ß) atoms of the same tyrosine residue. The
corresponding C(a)H is found at 5.33 ppm which shows a strong NOE connectivity
with one of the C(ß) protons (2.63 ppm) and a weak one with the other C(ß) proton
(3.12 ppm). The assignments are confirmed by the fact that these resonances are
absent in Fig. 3 indicating the close distance to the flavin. The assignments are
summarized in Table 1. Again unfortunately, no COSY connectivity could be detected
between C(a)H and the C(ß)H2- The dihedral angles of the side chain can be esti
mated from the NOESY intraresidue connectivities, i.e. xi must be close to 180°
or 60°, and x2 1S likely close to 0°.
The resonances due to Ala-56 have been assigned above, based on the results
of Figure 1 (Table 1). The intraresidue NOE connectivities for this residue are
also clearly observed in Figure 4, supporting the assignments and the conclusion
that this residue is located close to the flavin. Interestingly, Ala-56 is also
very close to the above mentioned tyrosine residue as clear connectivities are
found between the ß-methyl group of Ala-56 (0.11 ppm) and the C(a)H of the tyro
sine residue. In addition a weak connectivity is found between the ß-methyl of
Ala-56 and the C(2,6) protons of the tyrosine residue. Combining these observa
tions and the crystal!ographic data on the related flavodoxin from Clostridum MP
it is justified to assign these resonances to Tyr-89.
Having assigned three spin systems to specific amino acid residues in the
immediate vicinity of the flavin, one would like to give the sequential backbone
- 178 -
assignments, as established by Wüthrich et a K (1982). Although several so-called
d,, d? and d, connectivities can be observed in Figure 4, the analyses are hampered
by the fact that almost no clear COSY connectivities could be established for more
complicated spin systems, i.e. when two nonequivalent C(ß) protons are present, as
shown by Nagayama and Wüthrich (1981). In some cases some probable connectivities
could be determined from NOE effects, but assignments to specific amino acid resi
dues would be too speculative and are, therefore, not further explored in this
paper.
Table 1. Assignments of resonances observed in two-dimensional H NMR spectra
of M.elsdenii flavodoxin in the oxidized and reduced state.
Residue Assignment and chemical shift in ppm
Atom(s) Reduced Oxidized
Ala-56 NH 6.47 6.31
a-CH 2.66 2.32
g-CH3 0.11 -0.04
Leu-62 6-CH3 -0.65 -0.74
6-CH3 -0.23 -0.23
Y-CH 1.06 1.10
Tyr-89 NH n.o. n.o.
a-CH 5.33 5.26
f3-CH2 2.63, 3.12 2.90, 3.14
C(2,6)H;, 7.06, 7.42 7.06 2
C(3,5)H2 6.59 6.66
Trp-91 NH n.o. n.o.
a-CH 5.04 4.88
ß-CH2 3.28, 3.50 3.38
C(2)H 7.08 7.09
C(4)H 7.53 7.61
C(5)H 7.10 6.80
C(6)H 6.94 5.57
C(7)H 7.79 6.41
Isoalloxazine N(5)H 5.63
C(6)H 6.50 n.o.
n.o. = not observed.
- 179 -
On the basis of the crystal structure of the related flavodoxin from Clostri
dium MP (Burnett et al_., 1974) the resonance at -0.65 ppm was already assigned to
one of the C(6) methyl groups of Leu-62 (Moonen and Mülller, 1983b). No safe
assignments could be established for the whole side chain of Leu-62 using the com
bined NOESY and COSY results. The resonance at -0.65 ppm however gives a clear NOE
connectivity to the singlet at 6.38 ppm, which in turn shows a NOE connectivity to
the C(a)H of Ala-56, indicating a small distance between the methyl group of Leu-62
and the ß-CH, of Ala-56. To arrive at reliable sequential assignments, additional
experiments are needed, such as double and triple quantum coherence experiments 2
of flavodoxin in pure H^O.
9xldlzed_flayodoxin
The assignments of resonances to Ala-56, Tyr-89 and Trp-91 in the hydroquinone
state of the protein facilitate the assignments of resonances to these amino acid
residues in the oxidized state of the protein, because most resonances are fortu
nately well resolved. The COSY-spectrum of oxidized flavodoxin shows the complete
spin system of Ala-56 (Table 1 ) , although the COSY connectivity between the C(a)H
and the NH is considerably weaker than in the hydroquinone state. The connectivi
ties are also clearly observed in the NOESY spectrum (Figure 5 ) . The S-ChU of
Ala-56 shows in addition a NOE connectivity to a resonance at 5.26 and a doublet
at 7.06 ppm. This doublet belongs to Tyr-89 (Table 1). The resonance at 5.26 ppm
shows a strong NOE connectivity with the resonance at 2.90 ppm. This in turn ex
hibits strong NOE connectivities with the lines at 7.06 and 3.14 ppm. A strong
NOE connectivity is also present between the lines at 3.14 and 7.06 ppm. Therefore,
the resonances at 2.90 and 3.14 ppm are the two non-equivalent C(ß)H atoms and the
resonance at 7.06 probably represents the two equivalent C(2,6)H atoms of Tyr-89.
The C(3,5)H atoms are easily identified from the strong NOE and COSY connectivities
between the resonance at 7.06 and 6.66 ppm. Thus the C(3,5)H atoms are also equiva
lent. Weak NOE connectivities are found between the C(ß)H atoms and the doublet
at 6.66 ppm (Table 1). No COSY connectivity was found between the C(a)H and one
of the C(ß)H atoms in the quinone state either, as already discussed for the hydro
quinone state.
180
0
2
k
6
8
10
, ' | « ' < " • " * " " " * * • ' • « < - ' " - ' - • « • ; ; - *
• • * ' " . . . : ' " - • . < • ' . • / • • ; . : - . • • » • " . . * •
c * * o • • j ^ » . •* i n • * • - * o o o »- fff • • • • » • * ' - " ^ . •
.c ° . . " ° l \X"" / * , . •J*i"*-V' • •"*" °*- ••* •'• A '
»jgr * • * n % f * ' ^ * * "
• . d ^ ° * • • • • Ô. • < % • ' • • • ° ° . , o O • * ° . j ^ • •« . f i ». j,b o o t? O r <^. . * . • • « . .O 0- «
Of o • * O O O o o
i y « o • •
• • - T
y?. 0
t ' ' ' l 1 ! .
,* r { 0
r; r «
* V
• .•_
' , 0 m
\x+ t
' t '
a - -
10 6 U PPMIÔ)
Figure 5: Pure absorption (contour) plot of a H- H NOESY spectrum of the oxidized state of M.elsdenii flavodoxin in 0.1 M potassium phosphate buffer (pH 8.3), solvent consists of 90% H„0 and 10% H?0. Mixing time xM 0.1 s. For further details, see Materials and Methods. was
The complete spin system of Trp-91 has been assigned previously using another
technique (Moonen et eH., 1982b). The assignments are confirmed by both N0E and
COSY connectivities. Here again no COSY connectivity was observed between the C(o)H
and one of the C(ß)H atoms of Trp-91.
It is now possible to discuss the changes in structure and mobility upon re
duction for the three assigned residues and a detailed comparison is allowed with
- 181 -
the crystallographic results of the related flavodoxin form Clostridium MP
(Burnett et al_., 1974; Smith et a K , 1977; Ludwig et a]_., 1976). It has already
been shown (Moonen et al_., 1982) that Trp-91 in the quinone state is immobilized
and that its conformation in solution is identical to that described for the crystal
structure of Clostridium MP flavodoxin. This is confirmed in this study. The strong
NOE connectivities in the hydroquinone state show that Trp-91 is also immobile.
For Clostridium MP flavodoxin (Burnett et a K , 1974) it has been demonstrated that
the C(7)H atom of Trp-90 is the nearest atom to the N(5) atom of the flavin. This
also holds for Trp-91 in M.elsdenii flavodoxin in solution, as evidenced by the NOE
connectivity between N(5)H and C(7)H of Trp-91. The chemical shift of the different
atoms of Trp-91 indicates their relative position with respect to the flavin, as
deduced previously (Moonen et al_., 1982). However, it should be noted that the ring
current effect of flavin in the hydroquinone state (especially of the pyrazine sub-
nucleus) is decreased as compared with that of the oxidized state. This fact should
be considered when comparing the chemical shifts of the two states. Nevertheless,
the C(7)H of Trp-91 differs by 1.38 ppm in the two redox states, i.e. a strong
upfield shift occurs in the oxidized state and a small downfield shift in the hydro
quinone state as compared to that of free tryptophan. This shows that the atom in
question must be right above the isoalloxazine ring in the oxidized state and some
what displaced in the hydroquinone state, i.e. towards the N(5) atom of flavin as
deduced from NOE effects. The combined results show that the indole ring of Trp-91
moves towards the N(5) atom of flavin on reduction but the angle between the indole
ring and the isoalloxazine ring is hardly affected. The observed displacement, the
relative position of the two planes and the side chain dihedral angles are in excel
lent agreement with the crystallographic data of Clostridium MP flavodoxin (Smith
et al_., 1977). In the latter study it was pointed out that Trp-90 is particularly
susceptable to the influence of intermolecular contacts. Nevertheless the results
presented in this paper give strong evidence that the position of Trp-91 in M.els-
denii flavodoxin in solution is identical to Trp-90 in crystallized MP Clostridium
flavodoxin.
In both redox states the NOE connectivity found between the NH and the C(ß)
protons of Ala-56 is in accordance with the crystal structure of Clostridium MP
- 182 -
flavodoxin. This is also true for the NOE connectivity between the C(a)H of Tyr-89
and the C(ß) protons of Ala-56. However, the side chain dihedral angle of Tyr-89
differs from that of Tyr-90 in Clostridium MP flavodoxin. A dihedral angle xi of
about 60° is predicted for Tyr-89 from the crystal structure of MP Clostridium fla
vodoxin. This implicates that the NOE connectivity between the C(a)H and both C(ß)H
atoms should be nearly identical. Our study demonstrates, however, that one connec
tivity is very strong and the other is absent, indicating a dihedral angle of 180
or -60°. A similar analysis shows that x? is probably 0°, in contrast to the value
of ±90° in the crystal structure. Moreover the NOE connectivity between the C(ß)
protons of Ala-56 and one or both of the C(2,6)H atoms of Tyr-89 in the hydroquinone
state is not obvious, considering the crystal structure of Clostridium MP. Our study
indicates that the relative position of Ala-56 and the side chain of Tyr-89 differ
from that in crystals of Clostridium MP flavodoxin. In addition, we have shown that
the relative position of the two residues are identical in the quinone and the hydro
quinone state, but the rotational motion around the C(ß)-C(y) axis differs in the
two redox states. Equivalent C(3,5)H atoms are found in the quinone state and non-
-equivalent ones in the hydroquinone state. This indicates that the "flipping" rate
in the quinone state must be much faster than in the hydroquinone state. As strong
NOE connectivities are found between one of the C(ß) protons and the C(2,6)H atoms
3 9 the frequency of the rotational motion must be within the range of 10 and 10 Hz 3
for the quinone, and <<10 Hz for the hydroquinone state.
Although incomplete due to the absence of several C(a)H and C(ß)H2 COSY connec
tivities it is demonstrated that two-dimensional NMR experiments allow a detailed
analysis of some residues in the active center of M.elsdenii flavodoxin. Besides
several close agreements between the crystal structure of Clostridium MP flavodoxin
and the solution structure of M.elsdenii flavodoxin, some small distinct differences
are found. An extension of this study, using other two-dimensional NMR experiments,
is currently under active study in our laboratory and the results will be reported
elsewhere in the near future.
183 -
ACKNOWLEDGEMENTS
We are indebted to Mrs. J.C. Toppenberg-Fang for typing the manuscript, to
Mr. M.M. Bouwmans for the preparation of the figures and to Dr. C.A.H. Rasmussen
for carefully reading the manuscript.
REFERENCES
Anil Kumar, Ernst, R.R. and Wüthrich, K. (1980) Biochim.Biophys.Res.Commun. 95, 1-6. Billeter, M., Braun, W. and Wüthrich, K (1982) J.Bol.Biol. 55, 321-34?: Burnett, R.M., Darling, G.D., Kendall, D.S., LeQuesne, M.E.7~Mayhew, S.G., Smith,
W.W. and Ludwig, M.L. (1974) J.Biol.Chem. 249, 4383-4392. Ludwig, M.L., Burnett, R.M., Darling, G.D., Jordan, G.D., Kendall, D.R. and Smith,
W.W. (1976) in: Flavins and Flavoproteins (Singer, T.P., Ed.) Elsevier Scientific Publishing Co., Amsterdam, pp. 393-403.
Mayhew, S.G. and Massey, V. (1969) J.Biol.Chem. 244, 794-802. Mayhew, S.G. and Ludwig, M.L. (1975) Enzymes, 3r3Td. 12, pp.57-118. Mayhew, S.G. (1978) Eur.J.Biochem. 85, 535-547. Moonen, C.T.W. and Müller, F. (198217 Biochemistry 21, 408-414. Moonen, C.T.W., Höre, P.J., Müller, F., Kaptein, R. and Mayhew, S.G. (1982b)
FEBS Lett. 149, 141-146. and Müller, F. (1983a), Eur.J.Biochem., submitted, and Müller, F. (1983b), Eur.J.Biochem., submitted.
in: Current Topics in Chemistry: Radicals in Biochemistry
Moonen, C.T.W Moonen, C.T.W Müller, F. (1983)
108, pp. 71-107 Nagayama, K., Anil Kumar, Wüthrich, K
40, 321-334. Nagayama, K. and Wüthrich, Smith, W.W., Burnett, R.M.
117, 195-225. States, D.J., Haberkorn, R Steinmetz, W.E., Moonen, C
F.H. and Wüthrich, K. Tanaka, M., Haniu, M., Yasunobu, KTT
J.Biol.Chem. 248, 4354-4366. Tanaka, M., Haniu, M., Yasunobu, K.T
J.Biol.Chem. 249, 4397.
and Ernst, R.R. (1980). J.Magn.Reson.
K. (1981) Eur.J.Biochem. 114, 365-374. Darling, G.D. and LudwigTTT.L. (1977) J.Mol.Biol,
A. and Ruben, D.J. (1982) J.Magn.Reson. 48, 286-292. T.W., Anil Kumar, Lasdunski, M., Visser , T . , Carlsson, (1981) Eur.J.Biochem. 120, 467-475.
Mayhew, S.GTand Massey, V. (1973).
Mayhew, S.G. and Massey, V. (1974)
Van Schagen, C G . änd" Müller, F. (1981) FEBS Lett. 136, 75-79. Wagner, G. and Wüthrich, K. (1982) J .Mol.Biol. 155,~54~7-366. Wider, G., Lee, K.H. and Wüthrich, K. (1982) J.MôT.Biol 155, 367-388. Wüthrich, K., Wider, G., Wagner, G. and Braun, W. (1982) J.Mol.Biol. 155,
311-319.
- 184 -
Summary
H, C, N and P NMR was applied to a study of free and protein-bound
flavins in order to obtain a better insight into the mechanism by which the
function of the flavin coenzyme is „tuned" upon binding to apoflavoproteins.
13 15 A thorough C and N investigation of free flavins provided a detailed view
on the effects of sterical hindrance and polarization on the structure of oxi
dized and reduced flavins. Moreover, this study resulted in a semi-empirical
basis for the interpretation of NMR data obtained on protein-bound flavins.
Remarkable results of the study of free flavins are the fact, that oxidized
isoalloxazine is not fully planar in an apolar medium, and that fully reduced
isoalloxazine is more planar than believed up till now. Unsubstituted reduced 2
isoalloxazine in water has an almost fully sp hybridized N(10) and an N(5)
?
atom which is approximately 70% sp hybridized. Upon modification the hybri
dization of N(5) and N(10) can be modulated rather independently due to steric
hindrance. The so-called "butterfly" motion has a low activation barrier,
which is probably an important feature for the mentioned tuning mechanism.
Flavodoxin from Megasphaera elsdenii is the protein which is most exten-
13 sively studied in this thesis. By a C relaxation study it was shown that the
isoalloxazine ring is strongly immobilized upon binding to apoflavodoxin. Also
the phosphate group of FMN is strongly immobilized in the interior of the pro
tein. Moreover, the phosphate group is dianionic in the complex, regardless of
the redox state of bound FMN. Based on the NMR data, it was shown that the
redox potential for the transition semi quinone/hydroquinone is mainly governed
by charge interactions. For the first time a quantification of the redox poten
tial modification by the apoprotein could be established, thereby showing that,
contrary to literature reports, the redox potential modulation is not governed
31 1 by steric effects exerted on the isoalloxazine moiety. Both by P NMR and H
NMR it was shown that M.elsdenii flavodoxin acts as a one-electron transfer
ring protein shuttling between the semi qui none and hydroquinone state, because
the oxidized state is ruled out as a biologically relevant redox state in this
protein due to the introduction of a high activation barrier between the oxi-
- 185 -
dized and semiquinone state. Moreover, the electron transfer mechanism is of
an outer sphere type. Based on the NMR data some arguments are presented that
a specific complex formation between M.elsdenii flavodoxin and the electron
donor or acceptor is not needed for an effective electron transfer. An impor
tant part of the active center of M.elsdenii flavodoxin was elucidated using
time resolved photochemically induced dynamic nuclear polarization (CIDNP) and
modern two-dimensional NMR techniques. The use of the paramagnetic semiquinone
state in combination with 2D NMR techniques allowed the generation of NMR spectra
of only the active center. Complete assignments were given in both the oxidized
and the hydroquinone state for Trp-91, Ala-56 and Tyr-89, which are all very
close to the isoalloxazine ring. It was shown that the relative position of
Trp-91 with respect to the prosthetic group is slightly different in the two
redox states. The active center of flavodoxin from M.elsdenii and Clostridium
MP appear to be similar. The only small difference between results of the crystal-
lographic study on Clostridium MP flavodoxin and H NMR results on M.elsdenii
flavodoxin was observed for the relative position of Ala-56 and Tyr-89.
13 15 A detailed C and N NMR study was performed on the complex of riboflavin
and Riboflavin Binding Protein from the egg yolk and egg white. Subtle informa
tion on hydrogen bonding, conformation of the isoalloxazine ring and solvent
accessibility were obtained in oxidized and reduced state. As far as the results
could be compared with reported binding studies of riboflavin analogues, the
results appear to be in excellent agreement. The pyrimidine ring is exposed to
solvent, except for 0(4a) in both redox states. N(10) is forced into the molec
ular plane, which in turn probably causes N(l) to be somewhat out of the plane.
The NMR results are consistent with a partial opening of the protein at pH 6
and a more stable conformation at pH 9. Based on the NMR results a possible
function for the complex in the embryonic development is suggested.
As an example of the class of dehydrogenases, lipoamide dehydrogenase from
Azotobacter vinelandii was studied by C NMR. In the oxidized state hydrogen
bonds exist to 0(2a) and 0(4a), but the polarization of the isoalloxazine ring
probably does not extend to the N(10) atom. The 4 electron reduced protein
- 186 -
(both the essential disulfide and the flavin are reduced) contains an essen
tially planar N(10) atom, and again hydrogen bonds to 0(2a) and 0(4a). The 2
electron reduced protein (EHL) appeared to be particularly interesting. It
consists of an equal mixture of protein in which the disulfide is reduced
and the flavin oxidized, and protein in which the disulfide is oxidized and
the flavin reduced. The results allowed a detailed description of the elec
tronic structure of the two-electron reduced protein. From the NMR results
it is concluded that the redox potential of both centers is roughly the same.
The results show that if the disulfide is reduced one thiol group is present
as an anionic thiol group and extremely close to the C(4a) atom of the flavin.
The exchange of reduction equivalents between the two redox centers in EHp is
slow (<5 s~ ). In combination with reported kinetic data, it became evident
that NAD+ accelerates considerably the exchange of reduction equivalents be
tween the disulfide and the flavin. The results show that this transfer must
be implemented in the reaction cycle and offer a nice explanation of some
„anomalous" kinetic data.
As an example of the class of hydroxylases, p-hydroxybenzoate hydroxylase
13 from Pseudomonas fluorescens was studied by C NMR. Without substrate the iso-
alloxazine ring is probably exposed to solvent. Upon substrate binding a
drastic change of the polarization of C(2) occurs, which is in accord with
the published enzyme-substrate complex as revealed by crystallographic methods.
It is suggested, that the active center without substrate is rather mobile in
order to facilitate the binding of substrate. A conformational change accompa
nies the binding of substrate. The active center, as present in the oxidized
enzyme-substrate complex, is probably already essentially formed upon reduc
tion of the enzyme in the absence of substrate, which explains the drastically
decreased binding rate of substrate after reduction. Upon complex formation
with substrate in the reduced state the N(5) is forced into the molecular plane 2
of flavin. Some arguments are presented suggesting that the sp hybridization
degree of N(5) and N(10) are the main factors which govern the reactivity
(or „activation") of oxygen, but not the IT electron density at the C(4a) center
- 187 -
of flavin. An electron transfer from the hydroquinone state towards oxygen
probably precedes the formation of a C(4a) peroxyflavin by radical pair com
bination.
- 188 -
Samenvatting
'''H, 1 3 C , N en P NMR werd toegepast op een studie van flavines,
vrij in oplossing en gebonden aan apoflavoëiwitten. Het doel van deze studie
was het verkrijgen van meer inzicht in het mechanisme, waarmee de funktie van
de flavine co'ënzymen door binding aan de apoflavoëiwitten a.h.w. wordt afge-
13 15 stemd op de uiteindelijke funktie in het eiwit. Een diepgaande C en N
studie van vrije flavines leidde tot een grondig inzicht in de effekten van
sterische hindering en polarisatie op de struktuur van geoxideerde en geredu
ceerde flavines. Bovendien voorzag deze studie in de behoefte aan een semi-
empirische theorie ter verklaring van NMR gegevens verkregen van flavines
gebonden aan eiwitten. Opmerkelijke resultaten waren dat geoxideerde isoal-
loxazine niet volledig vlak is in een apolair medium en dat gereduceerd iso-
alloxazine in feite vlakker is dan tot dusverre werd aangenomen. Ongesubstitu-2
eerd gereduceerd isoalloxazine in water heeft een vrijwel sp gehybridiseerde
N(10), terwijl de N(5) ongeveer 70% sp gehybridiseerd is. Modifikatie blijkt
de hybridisatie van deze beide stikstof atomen onafhankelijk te kunnen beïn
vloeden d.m.v. sterische hindering. Bovendien werd duidelijk, dat de aktiver-
ingsenergie van de zgn. vlinderbeweging laag is, wat waarschijnlijk van belang
is voor het eerder genoemde „afstem" mechanisme van de flavin coënzymen.
Het flavodoxine van Megasphaera elsedenii is het meest grondig onder-
13 zochte eiwit in deze dissertatie. Door een C relaxatiestudie werd duidelijk,
dat de isoalloxazine ring stevig gebonden zit aan het eiwit en in feite ge-
immobiliseerd is. Ook de fosfaatgroep van FMN is sterk verankerd in het eiwit
en bovendien steeds dianionisch, onafhankelijk van de redox toestand. De NMR
resultaten maakten duidelijk, dat interaktie van ladingen de redox potentiaal
bepalen van de overgang semiquinon naar hydroquinon. Voor het eerst kon de
modifikatie van de redox potentiaal door het eiwit gequantificeerd worden.
Duidelijk is, dat eertijds gepubliceerde resultaten omtrent de oorzaken van
31 1
deze modifikatie onjuist zijn. Zowel P als H NMR toonden aan dat dit fla
vodoxine werkt als een één-electron overdragend eiwit, waarbij het eiwit pen
delt tussen de semiquinon en de hydroquinon toestand. Dit werkingsmechanisme
- 189 -
is gebaseerd op het feit, dat de geoxideerde toestand uitgeschakeld wordt
door de invoering van een hoge aktiven'ngsberg in the overgang tussen de
geoxideerde en semiquinon vorm. Het mechanisme van electronoverdracht is
waarschijnlijk van het „outer sphere" type. Bovendien lijken de resultaten
er op te wijzen, dat een specifieke binding tussen het flavodoxine en een
electron donor of acceptor niet nodig is voor een effektieve electronover
dracht. De struktuur van het aktieve centrum werd gedeeltelijk bepaald door
middel van tijd-opgeloste CIDNP (fotochemisch geïnduceerde dynamische kern
polarisatie) en moderne twee-dimensionale NMR technieken. Van groot belang
in deze studie bleek het gebruik van de paramagnetische semiquinon toestand.
Toekenningen werden gedaan voor Trp-91, Ala-56 en Tyr-89 in zowel de gere
duceerde als geoxideerde toestand. De drie aminozuren bevinden zich zeer
dicht bij de isoalloxazine ring. De relatieve positie van Trp-91 t.o.v. de
isoalloxazine ring verschilt enigszins in de twee genoemde redox toestanden.
Het aktieve centrum van dit eiwit gelijkt zeer sterk op dat van het verwante
flavodoxine uit Clostridium MP. De enige kleine verschillen tussen de rönt-
genstudies van het laatste eiwit en de NMR resultaten van M.elsdenii flavo
doxine werden waargenomen in de relatieve positie van Ala-56 en Tyr-89.
De interaktie tussen riboflavine en Riboflavine Bindend Eiwit werd
13 15 onderzocht d.m.v. C en N NMR. Subtiele informatie werd verkregen over
waterstofbruggen, conformatie van de isoalloxazine ring, en toegankelijkheid
van oplosmiddel. In zoverre de resultaten vergeleken konden worden met bind
ingsstudies van analoga van riboflavine, bleken de studies in zeer goede
overeenstemming. De pyrimidinering is blootgesteld aan water behalve 0(4a).
De N(10) wordt a.h.w. in het moleculaire vlak geduwd, terwijl N(l) enigszins
uit het vlak is. De resultaten zijn consistent met het idee, dat bij pH 6
het eiwit een meer open struktuur aanneemt dan bij pH 9. De beschreven infor
matie gaf aanleiding tot enkele suggesties omtrent de funktie van het complex
in de embryonale ontwikkeling.
Als voorbeeld uit de klasse van dehydrogenases, werd lipoamide dehydro-
13 genäse uit A.vinelandii bestudeerd m.b.v. C NMR. Waterstofbruggen naar
- 190 -
0(2a) en 0(4a) bestaan in het geoxideerde eiwit, maar de polarisatie reikt
niet tot N(10). In de 4 elektronen gereduceerde vorm, waarin zowel de
disulfide als het FAD gereduceerd zijn, is N(10) vrijwel vlak en opnieuw
bestaan er waterstofbruggen naar 0(2a) en 0(4a). Vooral de 2 electronen
gereduceerde toestant (EHL) bleek erg interessant te zijn. Het bleek te
bestaan uit een mengsel van enerzijds eiwit, waarin disulfide gereduceerd
en FAD geoxideerd is, en anderzijds eiwit, waarin disulfide geoxideerd en
FAD gereduceerd is. De electronische struktuur van de twee soorten werd be
schreven. De redox potentiaal van de beide redox groepen is vrijwel gelijk.
Indien disulfide is gereduceerd, dan is één thiolgroep anionisch. De over
dracht van elektronen tussen de twee redoxgroepen is langzaam (<5 s" ) .
Vergelijking met gerapporteerde kinetische data leerde, dat NAD deze over
dracht aanzienlijk versnelt. Deze kinetisch belangrijke stap moet toegevoegd
worden aan het reaktiemechanisme en verklaart enkele vreemde kinetische data.
Als voorbeeld uit de klasse van hydroxylases werd p-hydroxybenzoaat-
13 hydroxylase uit Pseudomonas fluorescens bestudeerd m.b.v. C NMR. Zonder
substraat is de isoalloxazine ring blootgesteld aan oplosmiddel. Na binding
van substraat treedt een verandering in polarisatie van C(2) op, wat in over
eenstemming is met het enzymsubstraat complex zoals dat gevonden is m.b.v.
röntgenstudies. Zonder substraat lijkt het aktieve centrum mobiel om de
binding van substraat te vergemakkelijken. Het aktieve centrum, zoals aan
wezig in het ES complex is waarschijnlijk ook gevormd na reduktie van het
vrije enzym zonder substraat. Bij binding van het substraat in de gereduceerde
toestand wordt N(5) in het moleculaire vlak geduwd. Enkele argumenten worden
aangedragen, dat de sp hybridisatie van N(5) en N(10) de belangrijke faktoren
kunnen zijn die de reaktiviteit (of aktivering) van zuurstof bepalen en niet
de TT electronen dichtheid van het C(4a) atoom. Waarschijnlijk gaat een elec-
tronoverdracht van hydroquinon naar zuurstof vooraf aan de vorming van een
C(4a) peroxyflavin door combinatie van het radikaal paar.
- 191 -
List of Abbreviations
Ac.rf CIDNP COSY CSA 2DNMR EDTA FAD FID FMN 2'FMN 3'FMN 4'FMN FMNH" FMNH„ pHBr HOMO Lipdh LUMO MeOH Me IA MeTMN MelMNH MelMNH MeLfl MeTARFH MeTARI Me9TARI MeTARIH Me?TARI NMR NOE NOESY RBP RBPW RBP Y TARF TARFH9 TMS ù
Tris TSP
2
tetraacetylri bof!avi n chemically induced dynamic nuclear polarization through bond correlated 2 DNMR spectroscopy chemical shift anisotropy two dimensional nuclear magnetic resonance ethylenediaminotetraacetic acid flavin adenine dinucleotide free induction decay riboflavin 5'-phosphate riboflavin 2'-phosphate riboflavin 3'-phosphate riboflavin 4'-phosphate anionic 1,5-dihydro-FMN neutral 1,5-dihydro-FMN para-hydroxybenzoate hydroxylase highest occupied molecular orbital lipoamide dehydrogenase lowest unoccupied molecular orbital methanol 3,7,10-trimethylisoalloxazine 7 methyl-10-ribity-isoalloxazine-5'-phosphate anionic 1,5-dihydro-MeIMN neutral 1,5-dihydro-MeIMN 3-methyl1 umi f1avi n 3-methyl-1,5-di hydrotetraacetylri bof1avi n 7-methyl-10-tetraacetylri bi tyl-i soal1oxazi ne 3,7-dimethyl-10-tetraacetylribitylisoalloxazine 1,5-dihydro-MeTARI l,5-dihydro-Me2TARI nuclear magnetic resonance nuclear Overhauser enhancement through space (NOE) correlated 2DNMR spectroscopy riboflavin binding protein riboflavin binding protein from egg white riboflavin binding protein from egg yolk tetraacetylri bof 1 avi n 1,5-di hydro-tetraacetylri bof 1 avi n tetramethylsilane tris (hydroxymethyl) aminomethane trimethylsilylpropionate lifetime
- 192 -
Curriculum Vitae
Chrit Moonen is geboren op 30 september 1955 in Stramproy. In 1973
werd het diploma gymnasium ß aan het Bisschoppelijk College te Weert
behaald. In hetzelfde jaar werd de studie Moleculaire Wetenschappen aan
de Landbouwhogeschool in Wageningen begonnen. Het doctoraal examen bevatte
een onderzoek op de vakgroep Virologie (3 mnd), een onderzoek op de vak
groep Biochemie (3 mnd), een onderzoek op de vakgroep Moleculaire Fysica
(hoofdvak), een onderzoek op de vakgroep Fysische Chemie (3 mnd) en een
tweede hoofdvak Biofysica (praktijk) aan de ETH in Zürich, onder leiding
van prof. Wüthrich. De studie Moleculaire Wetenschappen werd in januari
1980 cum laude afgerond. Vanaf eind januari 1980 werd aan deze dissertatie
gewerkt, in eerste instantie in dienst van de Stichting voor Zuiver Weten
schappelijk Onderzoek (Z.W.O.). Vanaf maart 1982 was hij in dienst van de
Landbouwhogeschool Wageningen. Sedert januari 1983 was hij in vaste dienst
aan de Landbouwhogeschool verbonden. Tijdens de werkzaamheden aan deze
dissertatie was hij betrokken bij onderwijs in het vak „Maatschappelijke
functies van de N-4 richtingen" en in het vak „enzymregulering".